Telco CDN. What potential for operators? M12113 December This study includes. - a report - a PowerPoint slide show

Transcription

1 Telco CDN What potential for operators? M12113 December 2012 This study includes - a report - a PowerPoint slide show

2 IDATE creates the DigiWorld Institute Founded in 1977, IDATE has gained a reputation as a leader in tracking telecom, Internet and media markets, thanks to the skills of its teams of specialized analysts. Now, with the support of close to 40 member companies which include many of the digital economy s most influential players the newly rebranded DigiWorld Institute has entered into a new stage of its development, structured around three main areas of activity: IDATE Research, an independent observatory whose task is to keep a close and continual watch on digital world industries, collect relevant data and provide benchmark analyses on market developments and innovations in the telecom, Internet and media sectors through its comprehensive collection of market reports and market watch services. IDATE Consulting, time-tested analysis and consultancy solutions: Our multi-disciplinary teams of economists and engineers established their credibility and independence through the hundreds of research and consulting assignments they perform every year on behalf of top industry players and public authorities. DigiWorld Institute, a European forum open on the world: The DigiWorld Institute will take existing IDATE initiatives, such as the DigiWorld Summit, the DigiWorld Yearbook and the monthly clubs in Paris, London and Brussels, to the next level. Members have the opportunity to participate in think tanks on the core issues that will shape the industry s future, drawing on the knowledge of outside experts and our own teams. Contributors > Vincent BONNEAU, Head of Internet Business Unit Vincent BONNEAU is the director of the Internet Business Unit, leading all the assignments and publications related to the Internet industry and on emerging technologies. His activities are especially focused on technological and marketing innovations. Vincent is an expert on technical and business aspects related to Net Neutrality and content delivery. Prior to IDATE, Vincent worked for the French Trade Commission (Economic Department of the Embassy of France) in San Francisco, USA as an analyst in charge of the software and web industries. He has also worked for marketing departments at several telecommunication companies including NOOS (French leading cable operator, now Numericable), WANADOO (now Orange) and FRANCE TELECOM. Vincent graduated from leading French engineering schools Ecole Polytechnique and from Ecole Nationale Supérieure des Télécommunications. He was also a visiting student with the class of MSc from HEC (top business school in France) in IT Management v.bonneau@idate.org > Samuel ROPERT, Senior Consultant > Tiana RAMAHANDRY, Senior Consultant ISBN ISSN Copyright IDATE 2012, BP 4167, Montpellier Cedex 5, France All rights reserved. None of the contents of this publication may be reproduced, stored in a retrieval system or transmitted in any form, including electronically, without the prior written permission of IDATE. IDATE and DigiWorld are the international registered trademarks of IDATE.

3 Contents 1. Executive Summary Arbitrages for telcos about telco CDN on handling growth in video traffic Confrontation in CDN between operators, traditional CDN and Internet giants A limited position for telco CDN in the short term, but with significant long-term potential Methodology Major telco CDN principles Context Global landscape Content distribution options on the Internet Traditional CDN Associated technologies Markets Telco CDN and transparent caching Principle of telco CDN Principle of transparent caching Telco CDN as a new business model for operators? Many objectives for telco CDN Player deployments and strategies Operator positioning AT&T BT Comcast Deutsche Telekom KDDI Orange SFR Verizon Telefonica Telecom Italia KPN KT Telstra TDF Bell Canada Telecom Argentina Equipment supplier positioning Summary Cisco Alcatel-Lucent Juniper Networks Huawei Ericsson IDATE

4 Nokia Siemens Networks HP CDN positioning Summary Akamai Limelight Networks Edgecast JetStream Transit provider positioning Summary Level Tata Cogent Major content provider positioning Summary Google Netflix Facebook Positioning of other technical service providers involved Summary SkyTide PeerApp Qwilt Verivue BlueCoat Strategic Analysis Regulatory issues and Net Neutrality Technical Issues CDN sizing What content can be cached? DNS redirection issue Problems implementing analysis tools Multiscreen and mobile QoS Federation of CDN CDN interconnection Broker concept Load balancing and multi-cdn management Developments for federation Economic issues Network investment Economic transfers between players Decisions on setting price Relationships between players Market Potential IDATE

5 Tables Table 1: CDN industry rates Table2: Telco CDN Initiatives Table 3: Acquisitions and partnerships of the major equipment suppliers Table 4: Equipment supplier positioning Table 5: Pricing based on customer location Table 6: Positioning of various traditional CDN around telco CDN Table7: Various positions of major transit providers Table 8: Various distribution choices for content providers Table 9: Primary transparent caching providers Table 10: List of the main CDN brokers Table11: Retail and wholesale revenue from video distribution *** Figures Figure 1: Change in Internet traffic by service category, Figure 2: Worldwide investment in fixed and mobile networks, Figure 3: Comparison of the increase in traffic costs and access revenue Figure 4: Share of Google's traffic using direct peering Figure 5: CDN concept Figure 6: Traffic simulation tools Figure 7: Akamai's mobile architecture for iphone Figure 8: Description of traffic optimisation Figure 9: Worldwide CDN market Figure 10: Changes in the share of value added services at Akamai, Figure 11: CDN unit cost decline (compared to the previous year) Figure 12: Principle of transparent caching: interception and redirection to a cache if the content is "cacheable" Figure 13: Changes in the revenue model for operators with telco CDN Figure 14: AT&T's CDN strategy Figure 15: Argument for CDN federation, Figure 16: BT's role in the value chain Figure 17: Various business models offered by BT Figure 18: Description of BT Content Connect with the various classes of service Figure 19: Comcast's road map Figure 20: Map of Comcast's CDN Figure 21: Map of the Edgecast "network" Figure 22: Evaluation of Deutsche Telekom's CDN implementation strategy Figure 23 : Creating a network dedicated to QoS Figure 24: CDN interconnection using an OCX platform Figure 25: KDDI Infrastructure Figure 26: Objective for acquiring CDNetworks Figure 27: Telco CDN used to converge KDDI's fixed and mobile networks Figure 28: Orange's CDN presence IDATE

6 Figure 29: SFR's distribution chain Figure 30: End to end CDN product Figure 31: Verizon's CDN coverage in the United States Figure 32: Telefonica's presence Figure 33: Telefonica's distribution network deployment project Figure 34: Telefonica's telco CDN as close as possible to the customer Figure 35: P4P Architecture with ISP intervening Figure 36: CDNetworks' global presence Figure 37: KT's data nodes presence Figure 38: KT's CDN deployment Figure 39: Telstra's CDN solution Figure 40: Targeted advertising using CDN Figure 41: TDF's offer (Media Services) in the value chain Figure 42: Evolution of CDN business models Figure 43: Bell Canada's CDN network Figure 44: Telecom Argentina's CDN architecture Figure 45: Structure of equipment suppliers CDN offers Figure 46: Cisco's Videoscape solution Figure 47: Cisco's CDS platform Figure 48: Ingestion, caching and distribution Figure 49: Cisco Mobile Videoscape Figure 50: CDN implementation in BT's network Figure 51: Phase 1 of the CDN operator federation pilot Figure 52: Phase 2 of the CDN operator federation pilot Figure 53: Velocix Digital Media Delivery platform Architecture Figure 54: Integration of CDN in the existing network Figure 55: Juniper's MediaFlow Controller Figure 56: Media Flow Solution position within the network Figure 57: Media Flow at the network edge Figure 58: Juniper's solution for mobile network Figure 59: Huawei's SmartCDN topology Figure 60: Location of icache in the network Figure 61: icache architecture, Huawei-Symantec transparent caching solution Figure 62: Ericsson's Media Delivery Network functionalities Figure 63: HP's SpeedVideo CDN global architecture Figure 64: Description of Akamai's portfolio of services Figure 65: Illustration of telco CDN deployment based on Akamai technology Figure 66: List of Akamai's offer for operators Figure 67: Extract of a presentation on CDN federation, May Figure 68: Types of applications targeted by the Ericsson-Akamai partnership Figure 69: Outline of the Ericsson-Akamai partnership Figure 70: Akamai's view on mobile: extend its presence to the foot of the tower Figure 71: Growth in Limelight Network's revenue and net profit Figure 72: Value added services' share in Limelight revenue Figure 73: Limelight Deploy description Figure 74: Edgecast's Network Figure 75: Edgecast and PeerApp partnership proposition Figure 76: Edgecast's Vision Figure 77: Telco CDN Federation Figure 78: JetStream's Streamzilla CDN network Figure 79: Description of Level3's services Figure 80: Description of Level3's "Intelligent Traffic Manager" IDATE

7 Figure 81: Change in various financial indicators Figure 82: Key metrics for the Level3-Global Crossing combination Figure 83: Effect of Level3's acquisition of Global Crossing Figure 84: Combined networks of Level3 and Global Crossing Figure 85: Tata Communications' product line Figure 86: Detailed description of Tata's CDN offer in India Figure 87: Change in Cogent's revenue Figure 88: Google's quarterly CAPEX spending since Figure 89: Distribution of costs for the various CDN Netflix uses, before (on the left) and after (on the right) introducing Open Connect Figure 90: Netflix's CDN peering points Figure 91: One day's traffic on Facebook in North America Figure 92: SkyTide role within CDN landscape Figure 93: Example of measures taken by SkyTide Figure 94: Federated CDN Figure 95: PeerApp's transparent caching solution Figure 96: Cache improves flow Figure 97: Amount of traffic that can be cached Figure 98: Qwilt's distribution solution Figure 99: Deployment of Verivue's tools Figure 100: How BlueCoat's transparent caching works Figure 101: How BlueCoat's CacheFlow transparent caching works Figure 102: Optimisation by consolidating servers Figure 103: Storage based on popularity Figure 104: Ability to cache based on position in the mobile network Figure 105: CDN interconnection concept Figure 106: Cisco's vision for CDN interconnection Figure 107: Federation of CDN initiations Figure 108: Illustration of the performance of various CDN on Deutsche Telekom's network Figure 109: Cascading model revenue flow Figure 110: Illustration of the "pay per cache" model Figure 111: Telco CDN market, IDATE

8 1. Executive Summary Key findings Telco CDN, or the deployment of a Content Delivery Network (CDN) in an access operator's network, is a major technical and economic response by operators faced with the growth in Internet traffic, especially video. Telco CDN is essentially being driven by the major incumbent operators (especially AT&T, BT, Orange, Telefonica, KPN and Verizon) with fixed video CDN. Operators can achieve several objectives with this solution: improve QoS, reduce network costs and traffic using caching servers and generate new revenue streams, especially from OTT players, while continuing to capture revenue from end users (through managed services) in a two sided approach to the market. Some of them can reduce deployment costs of some services using hybrid infrastructures. The operational positioning of telco CDN for operators is still relatively complex. There are several options, even though some of them have limited potential. However, transparent caching (which creates the equivalent of a telco CDN without agreements with OTT players) is growing in popularity as it is easier to implement. Actually, telco using telco CDN or caching have some decisions to make on setting the right price to attract enough traffic without setting a too aggressive price, as well as on revenue generated from managed services. Telco CDN is actually also a wholesale service which competes directly with operators' managed services for retail VOD. Telco CDN developments by operators can upset the value chain, especially for traditional CDN players. The arrival of operators introduces additional competition in an already very competitive market, which is also undergoing heavy consolidation. The threat of competitive confrontation with telco CDN leaves room for co-opetition, with all the traditional CDN still looking to sell CDN services to operators as basic reselling of production, technology licenses or outsourced services. CDN challengers (Edgecast, JetStream) were the first to position themselves like this with the major operators (Deutsche Telekom, KPN), joined in late 2011 by CDN leaders. Competition is stiff, and traditional CDN must adapt to interconnection imposed by the operators (especially paid peering). Operators compete mostly with the Internet giants (Google, Netflix), who have their own infrastructure (and even their own CDN), and who represent the majority of the traffic, and are not inclined to use telco CDN. Nevertheless, the impact of telco CDN is very small, due to the current lack of a federation that would provide larger coverage than the natural boundary of a telco CDN (i.e. more than only the operator's subscribers). Telco CDN will remain a very small market in the short term, with 4.7 million EUR in 2012 (or only 0.2% of the total market). But the opportunities in the medium and long term certainly exist, with enhanced offers beyond video, mobile development (in which traditional CDN are not competitive) and coverage of poorly served geographical regions. The potential for 2017 is estimated at 360 million EUR (or 5.5% of the total market). However, this forecast may be low, given that the forecast for 2020 is even larger (nearly 2 billion EUR in 2020). In addition, operators view this as strategic positioning in this market, as the benefits are much greater than direct profits (cost savings, improved QoS/QoE, ability to segment offers, two sided business model, etc.). IDATE

9 1.1. Arbitrages for telcos about telco CDN on handling growth in video traffic For telecom operators, optimising traffic, especially video, has become a major priority, especially during this telecommunications industry recession. There are several ways to position themselves: in paid peering, transparent caching, telco CDN (CDN deployed in the access network) as well as end user services (managed services, OTT services). Paid peering generates some revenue from the largest traffic generators, but it doesn't provide any real optimisation of the access network (only QoS is improved due to the shorter connections). In contrast, transparent caching solutions reduce traffic in the network, but don t involve any revenues from players upstream in the distribution chain (content providers, CDN). Unlike the two other solutions, telco CDN takes a two-sided approach to the market using a basic technique similar to transparent caching (generating revenue from both end users and OTT players) with an offer of better QoS, while optimising costs. Several operators have decided to follow this course, the most advanced ones now are Telefonica, Verizon and Orange working now with Akamai (even though the first ones were AT&T, BT and KPN, in addition to Deutsche Telekom's CDN reselling agreements). Their offers are essentially for CDN on fixed networks, with few value added services. However, operators must make several major decisions as they position themselves. The first one is what priority to give to each objective, between cost savings and generating revenue. Cost savings are often favoured, because they are based on the internal optimisation of the network, heavily dependent on the location of the servers in the network (currently mostly in the collection network). These deployments can also be financed by managed services as part of implementing managed OTT hybrid architectures. However, savings are easier to achieve using transparent caching than with CDN. Transparent caching is heavily deployed in geographical areas where copyright regulations are less stringent. In the case of telco CDN, contracts need to be signed with OTT players, who could refuse to sign for strategic reasons, avoiding having their traffic distributed by an operator who is also their competitor (especially in managed services). The second decision for operators wanting to generate revenue is on the price level for telco CDN. A low price attracts lots of traffic while a higher price would be uncompetitive (compared to traditional CDN) and would attract little traffic. Therefore, generating revenue will depend on the ability to attract traffic. However, it also impacts costs savings. Attracting less traffic implies achieving less cost savings. Finally, the last decision, sometimes not taken into consideration, is the operator's position, involved in both OTT distribution and managed distribution. Their retail business (selling VOD on managed networks) and wholesale (selling telco CDN to OTT players) are actually in competition with each other, each having very different margins based on the services sold. Retail generally offers better margins for operators (who get a share in the revenue, not just in the transit services, which are often more competitively priced) IDATE

10 1.2. Confrontation in CDN between operators, traditional CDN and Internet giants Telco CDN is obviously not the only offer in the CDN market. This market is already very saturated and competitive. The last few years have seen a significant wave of consolidation (notably with Akamai acquiring Cotendo) and a major drop in the price of CDN. This strongly limits CDN market growth, with "only" 24% growth in 2012 despite the explosion in traffic. In addition, in this CDN market of 2.1 billion EUR in 2012, video represented less than half of the market (45%) while value added services (all other content) already represented more than half the revenue. The arrival of operators in the CDN market is bad news for traditional CDN players. It intensifies competition and could potentially increase price pressure. In addition, operators could offer better QoS, being able to place cache servers lower in the network. This difference in performance has not really been proven yet, but could easily materialize in the future with connected TV and mobile services. The CDN market leaders (Akamai, Limelight) initially saw direct competition, while the CDN challengers (Edgecast, JetStream) quickly saw an opportunity. The CDN leaders have been imitating them since the summer. Most of the CDN offer technology licences for their solutions as well as managed services, helping operators develop their CDN faster without having to manage all the elements. This approach seems inevitable for CDN players, especially to eventually address the mobile market to avoid paid peering often applied to traditional CDN by the operators. Operator initiatives in CDN seem to be a way to attract major media players, who are used to paying for CDN services and to manage several CDN providers at the same time, using advanced load balancing tools to optimise this management (Cedexis). In contrast, relationships with the Internet giants seem more complicated. Most of them aren't interested in collaborating because they can live with the current level of QoS. Moreover, most of the major Internet players are deploying their own Internet infrastructure and operate a form of CDN (Google, Amazon and more recently Netflix) Finally, confrontation is partly strategic, with players wanting to avoid having providers (telco CDN) who are also their competitors (managed services). The giants will be difficult to attract, as well as small content and service providers who only work with a few suppliers. Actually, the biggest problem with telco CDN is its limited coverage. Traditional CDN can target all Internet users, while the telco CDN can only cover an operator s own subscribers. Only OTT players, already managing several CDN, can really consider going with a telco CDN directly. To respond to this coverage problem, operators are pushing for a federation approach, helped by telecom equipment suppliers (notably Cisco and Alcatel- Lucent) who have all gotten into the CDN market (often through acquiring companies). This is to create a "one-stop shopping" solution to access several ISP. The federation could even involve traditional CDN, playing the role of a broker. However, traditional CDN (at least the leaders) don't seem to be interested in this as it would turn them into a commodity. In addition, federation is hard to implement technically, to make cache servers interoperable as well as all the layers of value added services, which is a major part of the market. IDATE

11 1.3. A limited position for telco CDN in the short term, but with significant long-term potential The short-term opportunities for operators in CDN are somewhat limited, as their offers are generally incomplete. Telco CDN are focused on fixed video with few value added services and a limited boundary due to the lack of federation. There are few CDN in the more competitive CDN markets like the United States. AT&T revised its operations in 2011 and especially in the summer of 2012, staying in this market but using traditional CDN. Telco CDN only generated 4.7 million EUR in 2012, or barely 0.2% of the total CDN market. However, the long-term potential for telco CDN is significant. Operator CDN are looking to enhance their offers beyond just video, taking advantage of traditional CDN expertise and the entire telecom equipment suppliers' ecosystem. In addition, Implementing a type of mobile CDN is impossible without the operators because the cache servers need to be located as low in the network as possible (congestion being more on the last mile rather than in the cellular network. Cache servers are insufficient and mobile CDN must combine both video optimisation solutions with traffic management solutions, which generally are outside the competencies of traditional CDN. Traditional CDN can't act alone, including Akamai, which established a partnership with Ericson. Because of their local presence, operators are often better positioned to target certain geographical areas, also poorly served by traditional CDN. This is especially true in Asia, as well as throughout Europe, with mostly North American CDN leaders covering just a few cities. Telco CDN will accelerate the CDN market in certain segments and will capture market share in existing segments. It should represent about 5.5% of the entire market in 2017, or 360 million EUR, an average annual growth rate of 138%. Though this revenue is minuscule for operators, don't forget that at the same time they can achieve savings on transporting traffic and improved QoS. Even in the medium term, this market remains small for operators, because it will take time to effectively deploy into the mobile networks (LTE migration could help) and develop an efficient federation. In the long-term, i.e. by 2020, with heavy traffic from mobile, the opportunities will be even more significant (telco CDN could benefit from mobile to reach about 15% of the market by 2020 and more than 2 billion EUR). Operators will continue to monitor this market closely, whose significance is more strategic than financial. CDN is one of the essential foundations for positioning in a two-sided approach to the market (i.e. wholesale and retail) and segmenting offers by QoS. IDATE

12 2. Methodology The methods employed by IDATE s teams of analysts and consultants are based on an approach that combines: research and validation of data collected in the field; the application of classic industry and market analysis tools: segmentation, competition analysis, strategic strengths, modelling, assessment and forecasts ; the expertise of specialists who contribute their own analytical capabilities and those of their network of market analysts. More specifically, the tools employed by IDATE s teams are as follows: 1/ A multi-disciplinary team of full-time consultants, specialised by sector of activity IDATE s analyses are performed primarily by our in-house consultants, and very occasionally by freelance market analysts. This approach allows us to capitalise on our pool of expertise through teamwork, sharing knowledge, ideas, contacts, viewpoints and key data. Each report is drafted by a team of specialists, overseen by senior consultants with a proven track record in their field. 2/ Primary and secondary research IDATE reports and databases are compiled based on primary data obtained from one-onone interviews with the sector s decision-makers, and on secondary data which is established by cross-referencing public sources and external databases. 3/ An integrated information centre sustained by a number of tools and proprietary databases Over the past 30 years, IDATE has established working and data organization methods and proprietary databases that trace the central chapters in the history of our sectors of expertise. Companies: IDATE s in-house data service tracks the latest news and events to come out of the top telecom, Internet and media industry companies around the globe. Innovative firms and start-ups are monitored by the market experts in the different Practices. Markets: IDATE s databases are derived from rigorous processing of fundamental economic variables (GDP, investments, exchange rates, demographics, etc.) and their relation to decisive sector-specific and national elements (capex, national market dynamics, etc.). Technologies: IDATE s organization by Practice provides us with an efficient means of tracking innovation. IDATE s engineers ensure in-depth understanding of the changing shape of products and services and of the latest innovations in the marketplace. 4/ Contents of the published reports Each IDATE market report details the structures and issues at play in the market being examined, the decisive forces (technologies, regulation, consumption) and the players involved. Particular emphasis is given to market assessments and forecasts, as part of the central premise. All market reports are laid out in a clear and concise manner, and illustrated with tables and graphs of key market data and trends. The process of drafting of a market report includes the following stages: analysis of the information available in the in-house databases, and review of analyses performed in the recent past; IDATE

13 based on a preliminary segmentation and assessment of the market, and as part of an validated interview guide, analysts conduct interviews that enable them to validate working hypotheses; a market model is then established, making it possible to test the hypotheses that have an impact on the market s development, and validated by a new round of interviews; and, finally, the report s conclusions are debated with the team responsible for the project and with expert consultants from the various fields involved; a final proofreading and editing/revision process, prior to the production of the final version of the report which is delivered to the client. 5/ Methodology used in this study This study was completed based on: IDATE's expertise in CDN, through the various studies over the last five years in France and internationally, and its participation in the European project, FP7 OCEAN ( launched in 2010 and which will be complete in 2013: about twenty specific interviews with equipment suppliers, CDN service providers and operators deploying CDN. participation in specialized conferences (CDN World Summit, CDN Asia Summit, CDN Summit, Mobile Video Optimization) as well as the Mobile World Congress. IDATE

14 3. Major telco CDN principles In this section, we will review the technical and business aspects of Internet traffic distribution, which explains why operators are looking to offer CDN. We will then present the major principles associated with telco CDN, after a detailed review of fundamental technologies and CDN markets Context The review of telco CDN (or operator CDN) is heavily influenced by the technical and business implications underling the debate on Internet Neutrality and issues in traffic distribution. Beyond considerations based purely on principle (notably pluralism, freedom of speech and non-discrimination), the debates on Internet neutrality have uncovered several major dysfunctional points in the technical and economic balance of the Internet, especially on problems with Internet interconnection: the precarious economics of certain types of content and services (often only temporary at first), limiting the resources that could be used to finance the network and bandwidth. Revenue from content and service providers is completely out of synch with the traffic generated. Traffic is sometime paid by the content providers, by buying transit or CDN services to improve QoS. However, these markets are undergoing heavy competitive pressure, resulting in a substantial drop in unit prices. Revenue generated is often tiny compared to the traffic carried by the Internet. In addition, players are taking advantage of local overcapacity and/or technological innovations to lower prices. Having a CDN therefore indirectly provides a form of differentiation. Peering provides free exchange of traffic, normally for equal volumes of traffic. The Tier 1 operators are those that have access to the entire Internet using peering, and all the other players including ISP. Though peering resulted in the Internet's fast expansion, the expansion of video (and other very asymmetrical services) put in question the principle that let each one manage its own costs without billing the others. The development of unlimited traffic consumption rate formula in ISP offers logically generated higher consumption, making it difficult to maintain an economic balance. The existence of managed services, which secure a portion of the network financing, but through which operators increasingly offer services similar to the open Internet (or even OTT) but with better quality of service. There is a risk of distorting competition, especially if the ISP apply traffic management policies to these services for OTT. Telco CDN will be analysed in this universe of traffic distribution bringing together mainly transit, peering and CDN (relying itself on transit and peering) while considering the pressure on content provider revenue and the existence of managed service to provide similar services. Telco CDN is one of possible responses to handle these various developments. IDATE

15 Global landscape Rapid traffic growth but no operator revenue The Net Neutrality debate started with the worldwide growth in traffic generated by the consumption of video services over the Internet. According to Cisco, video traffic grew at a pace greater than 100% per year on fixed networks, and at a higher rate on mobile networks. In 2014, video will represent more than half the traffic on fixed networks. Figure 1: Change in Internet traffic by service category, Source: Cisco Consumption in Internet traffic has never stopped growing over the last few years. The current trend is universal: growth in the number of Internet users Even in developed countries, there is still a significant amount of growth in the number of users on the fixed Internet that has not yet come to full maturity. The potential is locally very significant on mobile Internet, whose growth is spectacular in most countries (mobile Internet is still used by a very limited fringe of population, except for in a few Asian countries). increase in the time spent online As users become more proficient in using the Internet, they increase the time spent on the Web and other Internet services. multiple uses and screens Users can access several services, each having different limitations in terms of quality of service. increase in bandwidth needs for certain applications especially for HD and 3D video (eventually including connected television) Simultaneous use Users multitask, which implies additional bandwidth. Deployment of next generation networks (fibre, LTE) New networks make it easier to use more demanding applications. The gradual expansion of connected television, in particular, will generate new traffic growth. The encoding required to reconstitute an image of sufficient quality for television is three times as much as for a PC, which is why video is one of the most important problems. IDATE

16 The increase in traffic requires sustained network investment To handle growth in traffic as well as to develop new services, telecommunications operators must invest heavily in their network capacity, especially in the access network. Some new investment will have to be made to develop new markets and to monitor infrastructure better. Figure 2: Worldwide investment in fixed and mobile networks, (Million EUR) Mobile Fixe Source: IDATE Due to the colossal investment required for next generation networks, operators now don't want to create traffic problems due to overcapacity. Financing networks is potentially in danger, even though the problem is much larger than that of Net Neutrality (especially for fibre and LTE). The operators' major issue is to handle the near exponential growth in traffic. Operators must actually increase the capacity of their networks, at all levels in the network (from the local loop to the Internet access network. Equipment costs are dropping, but not as fast as traffic is increasing. With traffic per subscriber growing strong, costs per subscriber are all going up. Eventually the operators' situation will be caught in a trap. Figure 3: Comparison of the increase in traffic costs and access revenue Source: Deutsche Telekom IDATE

17 Content distribution options on the Internet Key options The choices for distributing traffic on the Internet are heavily dependent on the technical competence of the players, as well as their size. For example, video content distribution on the open Internet is based on economies of scale. CDN isn't the only way to distribute Internet traffic. A decision has to be made between the following approaches: Transit: commercial agreement in which a provider pays for bandwidth to "access" (i.e. send traffic) the entire Internet. Transit services generally include agreements on the terms of quality of service, even though routing is done on a "best effort" basis (referring to penalties for poor performance). Transit services are purchases of bandwidth by flow, i.e. in Gbps, with a time based commitment The traditional system is based on billing at the "peak at 95th percentile" (i.e. nearly at the peak of traffic) and not on the actual volume of traffic. Rates heavily depend on traffic capacity and geographic area. Peering: relationship in which two players exchange traffic, generally for free. This exchange sometimes incurs costs, each one having to carry at least its own actual costs, especially for equipment, associated with exchange (routers, collocation, circuits, electricity. HVAC, etc.) for each peering point, generally amortized over three years. Sizing the peering points is based on equipment capacity (maximum traffic managed by the server for example) considering actual peak traffic. Content Delivery Network (CDN): solution for optimising traffic distribution on the Internet based on a set of distributed servers linked together over the Internet. CDN can offer better quality of service on the open Internet and reduces bandwidth costs. CDN offers are most often billed based on volume (i.e. the number of Gb or Tb) and not on bandwidth flow, generally under an annual contract and a contract for average monthly consumption (it is easy to revisit by extending the contract when consumption rises during the contract) and penalties for exceeding the consumption fee. Video service providers didn't take into consideration peak traffic problems. Issues with peering Agreements between the major transport network operators are often driven by the concept of peering (even though the majority is transit), to reduce costs and improve QoS (with a direct and shorter interconnection). Traffic exchanges at interconnection points don't result in billing between operators as long as the traffic is overall symmetrical. Unlike the main transport network operators, CDN operators, which generate significant traffic, have gradually interconnected with access network operators to take advantage of the peering mechanism (often free). In nearly every case, a content provider's traffic will initially pass through at least one transit operator (eventually through a CDN) before arriving at the access operator. The first transit operator (positioned at the first access point or first-mile) is most often paid. To deliver traffic to the Internet, without controlling the rest of the traffic route, this first operator will distribute this traffic to other transport operators, leaving transit up to these operators. However, somewhere along the chain, traffic will be passed to an important ("Tier 1", "Tier 1" regional) operator able to distribute this traffic by peering. The other network operators involved in the traffic route will then, to the greatest extent possible, continue to peer to distribute this traffic, which reduces quality of service, until the traffic arrives at the access operator, who distributes it to its customer. There is often a break in paying revenue for traffic along the traffic distribution value chain, possibly well upstream of the access operator. IDATE

18 The free peering model was implemented to avoid billing for exchanges, expected if it balanced out in both directions. Each provider retains the money from its customers and carries its own costs. This model normally works when the traffic is symmetrical. Asymmetrical traffic means that the two parties exchanging traffic must support costs for interconnection (servers, collocation, etc.), while only one party gets revenue for consumption of service. Asymmetrical traffic (and its contribution for financing networks) puts the peering model into question. The major service providers are looking for a way to expand their own infrastructure. In addition to transit services and servers, they are building their own infrastructure that they will interconnect to transport and access network operators. But this interconnection generates new problems: on one hand, the risk of saturating the interconnection points; on the other hand, asymmetrical traffic at the interconnection points (a video platform like YouTube, for example, generates much more traffic than it receives). Figure 4: Share of Google's traffic using direct peering Source: Arbor Faced with the growing difference in traffic symmetry at the interconnection node in the transport network, implementing paid peering could be a possibility. Operators would pay for the non-symmetrical portion of the traffic they generate. Between service providers (with transit offers) and traditional peering (free exchange), paid peering: can integrate major content providers into the overall Internet transport architecture; gives major content providers the ability to decide between operating costs (transit charges) and investment (interconnection in transport network nodes),......all while paying for the transport network more equitably. IDATE

19 3.2. Traditional CDN Associated technologies A Content Delivery Network (CDN) is an optimisation solution for distributing traffic on the Internet, using a set of distributed servers linked together over the Internet. Objective The concept of CDN brings Web content to the end user most often using local servers on the edge of the Internet (edge servers) and to ISP using bilateral agreements between CDN and ISP. Access operators are interested in this approach because it generally reduces the data packet transit time entering its network, as well as - to a lesser extent - the associated transit costs (eliminating transit costs). By moving the content closer, you can avoid longer transits when several users make the same request. Without CDN, the access operator must pay for transit to transit providers transporting the content provider's packets (unless the access operator is a Tier 1 itself). Using external CDN, like Akamai, the operator no longer has to pay the transit costs. These transit costs have become insignificant in developed countries except for Asia, but are major costs in other countries. For third parties, either content providers or operators, a CDN is the equivalent of a single Autonomous System (AS) (if it is a virtual operator aggregating various backbone networks), able to optimise routing with the shortest AS path. CDN can offer better quality of service on the open Internet and reduce bandwidth costs. They also guarantee better availability of service and content consumed by several users. They optimise content distribution, notably video, by distributing connection requests. Operational concepts Various techniques are used (generally in combination), like: caching (storage of popular content close to users) In this case, content is actually moved physically closer to the user. This is what most CDN have deployed. It is mainly used for video services. switching (traffic directed to a virtual router that directs traffic to the best server available, using the best route available among the several backbones used). This is a virtual setup. The content is still stored in the same place, but traffic distribution routes are continuously monitored to determine the fastest path at point in time t, bringing content closer to the user, even though it may need to pass through an intermediary. Switching is mostly used to accelerate traffic for services where latency is very critical, like e-commerce. traffic distribution (towards an underutilized or closer server) using the same concept, by separating the same flow (users requesting the same content) but for users located in the same final network. It avoids passing all the traffic along the same route. CDN players rely on several transit providers. IDATE

20 Figure 5: CDN concept Source: IDATE CDN players' infrastructures Servers are generally linked together to optimally transfer or replicate content (all servers don't contain all possible content). They are fed content (Web page, static object, multimedia content) either by default (for the most popular content or very local content), or on demand (content is transferred from a central server to the local server on the first request, and is thus available for subsequent requests). Most often, it waits for the number of requests to reach a threshold for a given content before storing it in the cache servers. Local servers are dispatched around the world and in the main global Web connection points (Europe, United States, Asia) and rely on several backbones. CDN players rarely have their own network infrastructure (except for a few players like Limelight and especially transit providers) and rely on third party infrastructure. Service and content providers can use, through paid services, CDN platforms that have shared infrastructure, without having to make major investments. Even though CDN improves quality of service by reducing the amount of traffic and optimising its distribution, it can do nothing for congestion in the access network (improvement only comes from the transport network. So they can't completely guarantee quality of service. However, they enter into quality of service agreements (SLA) using basic indicators (availability, etc.). CDN generally offer simulation tools to service and content providers to see the value added by CDN services, 1 using a Web monitor in real time and performance comparisons with public routes IDATE

21 Figure 6: Traffic simulation tools Source: Akamai CDN and cloud computing CDN architectures are not "cloud computing" architectures, which also rely on servers distributed in datacenters. However, though the basic concepts are the same, especially from an economic point of view (the customer rents infrastructure rather than developing its own), the key underlying technologies are very different: CDN use cache servers whose objective is to reduce network traffic. The savings achieved are first based on bandwidth, the on using shared servers. The main issue is increasing quality of service for content and popular Web page consumption (CDN is useless for individual content): cloud computing relies mainly on virtualizing servers allowing access to various applications without dedicated servers. The savings achieved are mostly on shared servers rather than on bandwidth. Cloud service consumption is not necessarily like CDN (not cacheable content consumed by several users). CDN mobile As described below, the concept of CDN remains a "fixed network concept" with decentralised servers (edge servers) in a fixed network, not a mobile network (for now). Caching (one of the main components of CDN service) is nearly non-existent on mobile networks. Caching servers offered by CDN providers on mobile networks use cache servers located on the fixed network (see the diagram below) and are located on the edge of the operator's access network core. IDATE

22 Figure 7: Akamai's mobile architecture for iphone Source: Akamai The primary characteristics of mobile CDN providers are based on: technologies that can solve the inherent problems in a mobile network such as traffic shaping and adaptive coding technologies (described above); technologies for adapting the content in real time, such as the picture (resolution), correct size and format, for all types of devices. Figure 8: Description of traffic optimisation Source: Ortiva Wireless IDATE

23 Markets Surprisingly, video traffic and large file transport only represent less than half of the CDN market in terms of sales generated. Most CDN revenue comes from dynamic traffic acceleration of small files including several small objects in terms of file size (video is considered a large object/file), that have much higher volume. Actually, dynamic traffic acceleration is the other major CDN functionality and service targeting secured sites (ecommerce, e-banking for example) where latency is critical. The market was estimated at a little more than 2 billion for 2012 worldwide, divided between video and non-video. It should grow rapidly in the next few years based on expanding video services, especially multi-screen as well as Internet services in general. The growth rate should average 24% per year between , reaching nearly 4 billion EUR in 2015, considering the initial impact of telco CDN (impact on mobile beyond 2015 remaining relatively low). Figure 9: Worldwide CDN market (Million EUR) Source: IDATE In addition, unlike other data transport players, CDN services often go beyond simple transport and sometimes include in their fees other downstream functionality (management, recoding, security, geo-blocking, etc.) and upstream (statistical measurement, advertising, etc.). The fee paying services include more than only traffic transport. This only represented slightly more than half of the revenue at the end of 2010 while it was nearly 65% in According to IDATE, this ratio could reach 70/30 by IDATE

24 Figure 10: Changes in the share of value added services at Akamai, Source: Akamai Competition and price The unit price of CDN has been going down for years. Changes in rates could be heavily impacted by telco CDN expansion. Operators could try to offer very low prices to attract or CDNise traffic. In any case, the arrival of operators should further intensify competition and contribute to lowering prices. Table 1: CDN industry rates (USD/GB) Amount High range Low range 50 TB TB TB TB Source: Blog Streamingmedia The diagram below shows changes in CDN unit costs over the last few years. Figure 11: 0% -5% -10% -15% -20% -25% -30% -35% -40% -45% -50% CDN unit cost decline (compared to the previous year) Source: IDATE IDATE

25 There are three main reasons for the drop in CDN rates over the next few years: drop in the cost of servers and connectivity (based mostly on transit) CDN players don't directly operate telecom networks (only the cache servers for most of them, notably Akamai, the leader) but rather buy connectivity (or peer). Their connectivity service is directly related to their transit cost (plus a margin). telco CDN's arrival They directly compete with traditional CDN, and could put pressure on prices to attract content providers to capture as much traffic as possible. The entire CDN market faces a drop in the price of transit to continue to capture traffic and revenue Telco CDN and transparent caching Principle of telco CDN An operator CDN (called a telco CDN from here on out) is, as the name indicates a content distribution network deployed by an operator, which interconnects at the edge of its core network (but inside of the access network), and is therefore basically different from traditional CDN. The idea is always to get the content as close to the end user as possible, which means that the operator must put caches as low in the network as possible in the IP equipment, at the boundary of the metro network or directly in the access network. Putting caches in the access network is still hard to do if the equipment is not necessarily IP (for example DSLAM and base stations). Like traditional CDN, the telco CDN relies on commercial contracts with content providers for their content to be cached. Operator CDN are currently targeting video traffic and transport of large files for cloud computing. Telcos are interested in deploying their own CDN to optimise content distribution. CDN also reduces backhaul costs. Telco CDN eliminates a large amount of costs associated with traditional CDN broadcasting services. Content is stored and issued from their own network rather than passing through the gauntlet of other networks and costly transit links. By integrating the content distribution value chain, telco CDN give operators the opportunity to develop new revenue streams based on their distribution capacity. In addition, by being closer to the end user - an ISP subscriber - telco CDN can normally deliver better quality of service than traditional CDN, especially for large files. Telcos have several options for locating servers in their networks to take advantage of a hierarchical content storage system based on popularity. Telcos will locate caches based on their objectives (network optimisation, bandwidth reduction, content distribution acceleration): the BRAS equipment interconnecting the ISP network and DSLAM in a DSL network, or network heads for a cable operator, to minimise traffic across the backbone and directly distribute content to the end user. This protects resources, reduces bandwidth consumption and lowers the associated costs. upstream of the exchange or metro points to accelerate content distribution. at the mobile LTE network core, to optimise traffic flow Principle of transparent caching Transparent caching is based on the same technical principles as telco CDN. Operators intercept, cache and distribute unmanaged content. The fundamental difference with CDN is the lack of commercial contracts with the content providers, and no commercial involvement by telcos (reducing the cost of selling). IDATE

26 Figure 12: Principle of transparent caching: interception and redirection to a cache if the content is "cacheable" Source: Verivue For a CDN, traffic is delivered directly and is easy to isolate. Specific caching rules can be set between the CDN and content providers. For transparent caching, traffic is analysed to determine content that can be stored on the cache services (basically popular content). Transparent caching lets telcos offload that amount of traffic going through their networks. The interest in this is to manage OTT flows, that consume a lot of bandwidth, thereby improving network usage and QoS. The operator using transparent caching receives no revenue. This process is still marginal in Europe because of content regulations. Without authorization from the copyright holder (which is what happens in CDN), content can only be stored temporarily. But the idea of "temporarily" is poorly defined and leaves room open for claims against ISP from copyright holders. If storage is only (very) temporary, the solution loses its interest (the objective being to cache). Transparent caching could be of interest in emerging countries, to help with limited connectivity Telco CDN as a new business model for operators? A new interconnection The Internet expanded globally based on relationships for transit and peering, and CDN solutions. The model generally used up to now was based on: payment by a user to the ISP; payment by the content provider to the transit provider or CDN (who then pays the transit provider); free peering to the greatest extent possible between the transit provider and the ISP (otherwise the ISP pays the transit provider for transit). Future changes in the Internet interconnection market could involve several movements: most of the potential changes (with several of the current trends enduring) are concentrated on new transfers of value to ISP from implementing more segmented and differentiated services. These transfers of value could come from downstream (retail market, from the user) and/or upstream (financing Internet and media players directly or indirectly through transit providers and CDN), and are no longer exclusive (several models could be implemented). These models are based on upstream or downstream payment for traffic based on: volume of traffic (or peaks in traffic), i.e. on the units of traffic directly consumed; quality of service associated with the traffic (probably also billed based on volume), with different rates based on quality; elements generally independent of traffic itself, as in packaged offers. IDATE

27 One approach among many Telco CDN is only one of the major approaches (the other being paid peering) to change operators' business models, especially in the upstream market. The other major option for the upstream market is paid peering, which generates revenue but doesn't involve changes in the network and may not really create better QoS (beyond the possible benefits of direct interconnection). Figure 13: Changes in the revenue model for operators with telco CDN Source: IDATE Many objectives for telco CDN Operators can achieve several objectives by developing these solutions. They can: optimise their networks and associated CAPEX Using transparent caching reduces the overall amount of traffic to transport (and thereby the transport costs), reducing saturation problems. However, operators need to find a balance in deploying cache servers. The closer to the user, the more traffic is reduced but more cache servers are required (increasing costs). generate additional revenue from content providers to use CDN instead of transit or paid peering. The increased revenue can be used to finance the operator's CDN itself and the network in general. Ensure convergence of OTT and managed content distribution architectures Managed services are often based on dedicated solutions. Implementing hybrid solutions may reduce costs. Telco CDN and transparent caching solutions could be complementary to distribute all kinds of content, managed or not, as economically as possible. Both of them could play a role in achieving savings in the network. IDATE

28 4. Player deployments and strategies In this section, we present the major initiatives of the main players involved in telco CDN developments over the entire value chain Operator positioning Transparent caching or telco CDN? All the major operators, essentially the incumbent operators, have implemented telco CDN solutions, especially in Europe and North America. These solutions are certainly rarer in challengers (SFR being one of the exceptions). These solutions are mostly limited to fixed networks, generally at the collection level. Deployments on mobile networks are rare and most often consist of interconnection with the network core. Efficiency is still limited and the benefits are only economical in countries where transit is still relatively costly (recall that CDN is used to avoid paying transit costs). In countries outside of Europe (where regulatory constraints are not as strong), transparent caching is used quite frequently. It is actually very effective and highly suited for local networks, quickly resulting in savings. Video centric with different models In most cases, deployments are focused only on video traffic as well as other types of huge files. Operator global deployment strategies are essentially the same, using different technical solutions. Some operators like Deutsche Telekom and Telecom Italia are happy with licensing from CDN players (reselling third party CDN products). Others like Bell Canada and KPN go farther, but still rely on CDN players, as part of CDN managed services operated by the CDN players. Some operators like Telefonica and Orange are developing their own CDN, relying mostly on equipment suppliers. Different issues depending on geographic areas The first operators to take a position on the subject were AT&T and BT. They seemed to have set the pace, while players such as Telefonica and Verizon (and to a lesser extent Orange) are now the most advanced. The case of AT&T is a perfect example, leaning towards reselling CDN. However, the American market is probably the most complex, due to the strong presences of traditional CDN players (most of whom are American anyway). This makes Europe the more attractive potential market. CAPEX vs. revenue The priority for operators seems to be above all on CAPEX savings even among those who expect to monetize services. Commercialization is difficult and the CDN market is already very competitive. In addition, operator offers only address a portion of the market (video), with few value added services, which actually represent the majority of the market. The operators still pursue their efforts, especially to eventually address mobile in which they have a distinct competitive advantage over traditional CDN, by combining CDN and Traffic Management. They can also offer other services like e-commerce. IDATE

29 Table2: Telco CDN Initiatives Player Region Type of content AT&T World All but focused on video BT United Kingdom All but focused on video Technical Partner Date Other elements In-house to start with 2011: Edgecast and Cotendo (a few technologies) 2012: resell CDN (Akamai and Limelight are cited) Cisco (CDN offer via BT Wholesale) End of (new CDN) 2010 BT Content Connect Comcast USA Video Internal Deutsche Telekom Europe Edgecast 2009 Reselling partnership TP Poland Mainly video Verivue 2008 Division of France Telecom Group Verizon World Mainly video Content management Technicolor and Motorola Network: Alcatel- Lucent and HP France Telecom (Orange) 2011 Member of the P4P working group 200+ datacenters in 20 countries "Verizon Digital Media Services" (VDMS) Europe Mainly video In-house 2011 Interconnection studies with FP7 OCEAN Partnership with Akamai (late November 2012) KPN NL Mainly video JetStream 2009 Member of several interconnection projects Orcon NZ Mainly video Alcatel-Lucent (Velocix) SFR France Innovative services such as smart home and e-health Not limited to video 2010 Early 2012 Tata Asia Mainly video BitGravity 2009 Telecom Italia Telecom NZ Italy Mainly video CDNetworks Sept New Zealand Mainly video Edgecast Telefonica Europe In-house Launched in 2011 TeliaSonera IC Sweden Gaming, with a CDN-like infrastructure Now HD Video Telstra Australia Multimedia content Cisco June 2012 Telus Canada Mainly video Edgecast Mid-2010 Source: IDATE Reselling partnership Member of the P4P working group IDATE

30 AT&T AT&T entered the CDN sector in 2008 and was one of the first telcos to invest in this market segment. Since 2007, it has invested 70 million USD in the construction of content distribution capacity. The platform, called ICDS, and the technology in general, came from internal research and development. The network deployed has about 400 gigabytes of capacity over its own CDN, with a total of 33 nodes around the world. Just like traditional CDN providers, AT&T also supplies value added services such as consumption information reporting (audience measurement, customised target advertising), content management (code conversion), etc. According to AT&T, in 2010, 30% of traffic transiting its core network was cacheable content for its customers. To meet time-to-market, AT&T had followed its strategy, starting at the end of 2010, based on internal development, with strategic partners Edgecast and Cotendo providing the technologies. In mid-2012, there were several rumors that they were changing their telco CDN strategy. AT&T were going to change to reselling CDN (Akamai and Limelight were cited as partners). These rumors appeared to confirm that AT&T was dropping out of internally developed telco CDN solutions. According to these rumors, a partnership with the service providers cited above would ensure minimum guaranteed revenue (despite the amount of traffic) over several years. Figure 14: AT&T's CDN strategy Source: AT&T AT&T is one of the drivers in favor of a CDN federation. Actually, AT&T saw the advantages for a Federation or CDN interconnection early on, as shown in the extract of the presentation below. IDATE

31 Figure 15: Argument for CDN federation, 2010 Source: AT&T BT BT was one of the first telcos to effectively launch a telco CDN offer. The service is provided by its BT Wholesale division. It primarily targets ISP. The United Kingdom saw its video traffic grow very fast over the last few years due to the success of acquiring BBC content (the United Kingdom's public content provider) called BBC iplayer. This trend should continue over the next few months with the arrival of hybrid Connected TV (YouView) in the summer of Figure 16: BT's role in the value chain Source: BT IDATE

32 In any event, BT has a relatively flexible offer in terms of: Business models (see the diagram below) Two types of competing models: - one based on the consumer, who would pay to receive better quality of service, either to the ISP directly (who would pay BT), or to the provider who would pay a portion to BT; - one based on the content provider, who would pay BT directly (Wholesale CDN as shown on the diagram) or the ISP who would then pay BT. Various classes of service (from "basic" to "premium"), providing quality of service (QoS) by implementing techniques to label packets as well as prioritisation. Figure 17: Various business models offered by BT Source: BT BT has often expressed its strong interest in the CDN federation, whose objective is the help access operators content distribution. This content includes video as well as gaming and software updates. Figure 18: Description of BT Content Connect with the various classes of service Source: BT By using BT (blue arrow), content providers choose their class of service (3) then store their content in a cache, using Wholesale Content Connect (WCC) Cache (BT's caching offer). IDATE

33 Comcast Comcast exercises minute control of its network, which has sometimes resulted in complaints of "abusive use" of its network infrastructure: several complaints about P2P traffic, which have led it to take technical measures to block traffic. Comcast has been warned by the FCC: complaints against Netflix (through Level3), in December 2010, that it hasn't paid for the true value of using its networks. Video consumption has exploded, both in its managed services and over the open Internet (complaint about Netflix). Its deployment of CDN infrastructure was completed at the end of Its primary objective was to optimise its network internally, to deliver the best QoS possible for its video offer called "Xfinity". The content offer is more than 500 Tb and the video infrastructure has more than 700 streaming servers. Consumption is nearly 350 million views per month. Comcast introduced a hierarchical architecture of servers, with 4 major cache centers (Atlanta, Philadelphia, Chicago and California), regional centers and up to 130 local servers (for the most popular content). Figure 19: Comcast's road map Source: Comcast Figure 20: Map of Comcast's CDN Source: Comcast IDATE

34 Deutsche Telekom Unlike other players who opted for internal development, Deutsche Telekom decided to resell an existing solution from Edgecast. In early 2009, Deutsche Telekom signed a reselling agreement with Edgecast while taking advantage of its technical expertise, as well as its already existing deployments (21 PoP distributed over 4 continents). Deutsche Telekom deployed "its" CDN 25% faster than the competition using this model. The main reason for implementing a telco CDN was the improvement of QoS. Figure 21: Map of the Edgecast "network" Source: Edgecast Time-to-market was the most important criteria for Deutsche Telekom. In addition, this partnership resulted in cost savings in the short term, as well as generating revenue from reselling. Figure 22: Evaluation of Deutsche Telekom's CDN implementation strategy Source: Deutsche Telekom Regarding the CDN federation, Deutsche Telekom has already clearly stated its strategy. As seen in the figure below, Deutsche Telekom is calling for an Open Carrier Exchange (OCX) platform that will be a mediation platform, acting as a broker (or exchange platform) 2. Its objective is clearly to provide QoS end to end. 2 See section on "CDN interconnection". IDATE

35 Figure 23 : Creating a network dedicated to QoS Source: Deutsche Telekom Figure 24: CDN interconnection using an OCX platform Source: Deutsche Telekom KDDI KDDI has just started in the CDN sector. KDDI acquired CDNetworks in 2011, one of the main CDN providers in the world and number 1 in Asia, for 167 million USD. CDNetworks has 130 storage sites located in 70 cities in 31 countries. The objective of this acquisition was to manage the uncontrolled growth in mobile traffic. KDDI has a very concentrated network, as shown in the figure below. It has two datacenters in Japan and two in the United States. Figure 25: KDDI Infrastructure Source: KDDI IDATE

36 Figure 26: Objective for acquiring CDNetworks Source: KDDI KDDI's telco CDN objectives include: reduce backhaul traffic (collection network), reduce response time and increase QoS. KDDI's other objective is to "relieve" its mobile network. Combining the mobile network and CDN infrastructure remains its biggest challenge. Figure 27: Telco CDN used to converge KDDI's fixed and mobile networks Source: KDDI KDDI primarily targets content providers with its telco CDN. KDDI expects to initially save costs, and then generate additional revenue. Regarding CDN federation, KDDI thinks that the business models, which are not yet defined, should be established on a case by case basis. CDN interconnection (technical solution) should also be determined on a case by case basis. It should use identical technologies to make it easy to interconnect (non-standardized protocols). KDDI has participated in several tests with SFR in France and Telecom Italia in Italy, with Cisco as the technical partner (see Cisco profile). IDATE

37 Orange Orange has already deployed a CDN in two of the countries where it has a presence, generally targeting PoP in highly populated areas (about 25% of PoP); In Poland, targeting its customers to distribute Web and video services, with CDN also being used for internal services; in France, to distribute huge files (such as video) internally. Extending to external customers (traditional TV players for VOD and catch up TV) is still under investigation. Orange is also looking for a "corporate shared infrastructure between the various countries in which Orange has a presence. Figure 28: Orange's CDN presence Source: FT Orange's goal is to offer the best content distribution solution based on its network coverage (both fixed and mobile) and be a partner for the rest of the world. CDN is seen as an enabler for "Orange Smart Networks services", both for content transport as well as for brokering services (marketplace). In addition, Orange already relies on transparent caching, especially in the Middle East and in Africa for mobile telecom operators. The objective is to reduce the cost to connect to the Internet. This seems to be already possible, but needs to be applied in small networks (otherwise CDN would be preferred). Orange's current CDN is based on Cisco technology and can process most of the current technologies and protocols used for delivering large files (such as software). Improvements are still needed for security and DRM, while developing streaming technologies (in particular Smooth Streaming and HTTP streaming) is underway. It is still hard to evaluate the technological competitiveness of the offer, because the traffic load is still limited (except for the enormous peak in Poland for the president's plane crash). Even though the objective is to sell CDN products directly to content providers, CDN is not part of Orange's wholesale catalogue. Orange can offer two competing products: IP Transit and the CDN service, by providing direct access on the ISP network, which is an advantage in terms of distribution quality, compare to traditional CDN pure players. IDATE

38 Orange's CDN products can also be sold to traditional CDN providers, which require interconnection offers. Orange has no solution available today, but is working with other operators. Finally, Orange is relying enormously on cloud computing, where it sees several synergies for providing CDN services SFR SFR launched its CDN in 2009 for internal use (distribution of its own services). Before this, the sfr.fr site was hosted by Akamai. Once complete, CDN usage was expanded to other services, both managed services and open Internet. The implementation of CDS was done to eliminate free peering, which had become too unbalanced (with the implementation of paid peering in 2011). As of early 2012, even though the offer is available, SFR still does not have all the reporting tools (audience measurement, marketing) needed but the offer should be completed and competitive before the end of In addition, SFR does not expect to use transparent caching. Figure 29: SFR's distribution chain 1 Broadcast traditionnel Plateformes DVB / Satellite Réseaux Broadcast garantis DVB STB DVB / Sat. Réseaux télécom multi / unicast 2 Services managés 3 CDN Services managés Services managés Peering Plateformes de services / CDN opérateurs Plateformes de services / CDN opérateurs Réseaux opérateurs multi / unicast garantis Réseaux opérateurs unicast best effort ADSL FTTH ADSL FTTH WiFi Box + STB 4 CDN Transit Internet best effort 3G Pur OTT Source: SFR Its infrastructure includes 12 PoP in France, with sharing for local traffic, including flexible caches. SFR has some technical issues such as consolidating logs for reporting, the rise in management orders (via API from API) as well as DNS routing. In terms of business model, SFR fully expects to achieve cost savings as well as generate revenue. Billing will be based on bandwidth consumed rather than volume (a widespread practice by traditional CDN providers). SFR is interested in an international interconnection of CDN operators but does not expect to extend its offer internationally. SFR is also open to an interconnection with traditional CDN, even though they are not exactly in favour of it. SFR has also worked on this process during a pilot with Cisco and KDDI. IDATE

39 Verizon In 2008, while AT&T was launching its CDN strategy, Verizon opted for a P4P strategy (in collaboration with Telefonica). It expects to optimise P2P flow by avoiding exchanges between shared pairs, connecting pairs in the same access network. In addition, transfer speed would be significantly improved by reducing the path length. Nevertheless, in 2011, Verizon took the initiative to launch its own CDN, investing a total of 370 million USD. Called "Verizon Digital Media Services", the product was developed in partnership with several companies: for content management: Technicolor and Motorola; for network management: Alcatel-Lucent and HP Figure 30: End to end CDN product Source: Verizon It is deployed internationally, with an architecture based on more than 200 datacenters distributed throughout 20 countries. Figure 31: Verizon's CDN coverage in the United States Source: Verizon Verizon has some doubts about a CDN federation, which according to Verizon, is much too complex to implement. The main problem is the diversity (and the fragmentation) of solutions (especially technical) proposed by providers and telcos. IDATE

40 Telefonica CDN offer As with its competitors, Telefonica decided to deploy its own CDN to monetize traffic by distributing content to users who are willing to pay. Telefonica launched its CDN offer in 2011 relying on the deployment of more than 40 CDN nodes covering a dozen countries, with an aggregate capacity of 600 Gbps. Initially available in Spain and Argentina, the offer will be expanded to other countries such as Chile, Peru, Brazil, Germany and Venezuela. The operator is also present in China based on its alliance with China Unicom. In 2012, the objective is to increase capacity to 2000 Gbps. Telefonica's CDN business is integrated into its Telefonica Digital, the operator's digital division that includes cloud computing, M2M and even advertising. Figure 32: Telefonica's presence Source: Telefonica Figure 33: Telefonica's distribution network deployment project Source: Telefonica IDATE

41 Telefonica's CDN solution was designed to distribute high quality video and download 3D and HD content, in any format required by the user. Telefonica's CDN uses HP servers and combines different sizes of servers to position the content distribution units at various points in the network to get as close as possible to the customer. Figure 34: Telefonica's telco CDN as close as possible to the customer Source: Telefonica Telefonica's proposal is based on four different offers: Web dynamics, for accelerating distribution services (faster loading of Web pages and applications from any type of end user equipment). Video services, for distributing streaming video, by continuous downloading or on demand, in any format while guaranteeing QoS; intelligent downloading for distributing large files (i.e. 3D and HD)'; private distribution, to optimise content transport within an organisation. In addition, Telefonica has developed its Apollo solution to monitor CDN statistics. P4P, Telefonica's hybrid CDN solution deployed? Telefonica developed internally a hybrid CDN solution for worldwide distribution using P4P architecture to optimise P2P flow transport. Telefonica, along with Verizon, was one of the pioneers in integrating the P4P topology in its CDN network. The hybrid CDN solution includes a centralised concept using apptrackers in a given area to control and implement pre-caching, enforcing rules and gather statistics (see diagram below) then a decentralised architecture for content distribution, dynamic allocation and cooperative streaming. Recently, Telefonica stopped referring to this part of its telco CDN; the use of this architecture is unclear. IDATE

42 Figure 35: P4P Architecture with ISP intervening Source: Telefonica Telecom Italia Telecom Italia has based its CDN on CDNetworks' solution (now KDDI's property) through its international subsidiary, Telekom Italia Sparkle, to offer CDN services, as part of a distribution agreement signed in Telecom Italia Sparkle resells CDNetworks' suite of services in the three regional targets where it is present, i.e. Europe, the Mediterranean and Latin America, taking advantage of its worldwide presence (see the map below), guaranteeing fast download time and shorter latency. Telecom Italia's offer, Content Acceleration Network, was created to help: companies manage downloading large files from a Web portal, media content providers that require high quality direct transmission, content on demand and streaming to distribute to their customers. companies who need instant access to information from their websites for a globally distributed audience. In addition, Telecom Italia relies on Cisco's CDN solution using the Content Delivery System - Internet Streaming (CDS-IS) platform to provide multiscreen video services. It also can distribute television channels directly and on demand content through a portal (yalp.it) on any device (telephone, computer, tablet, etc.). Currently, Cisco's CDS-IS supports the main Web formats for downloading and content broadcasting including Adobe Flash Player. Regarding the CDN federation, Telecom Italia is involved in Cisco's pilot work. IDATE

43 CDNetworks CDNetworks was established in 2000 in South Korea. In 2011, the Japanese operator KDDI acquired CDNetworks to manage the growth in its mobile traffic. As part of this acquisition, CDNetworks became a subsidiary of KDDI but continues to operate under its own name. CDNetworks is now one of the main CDN with more than 130 PoP distributed in 70 cities and 31 countries. CDNetworks is currently present in China with two datacenters. The company is positioned as an alternative to Akamai with a global capacity of 1,100 Gbps. Figure 36: CDNetworks' global presence Source: CDNetworks KPN In 2009, to respond to the strong demand from broadband subscribers and content providers to optimise online media content distribution, KPN decided to deploy its own CDN in its own network. KPN thought about developing a low cost solution internally. Instead, the operator decided to deploy its own CDN using a white label CDN solution based on JetStream VideoExchange technology. As a software licence holder, KPN was able to accelerate CDN service implementation, taking two months for deployment and two weeks for testing. During these two months, KPN took advantage of training provided by JetStream, integrated the solution and implemented functionality requests. According to JetStream, CDN of incumbent operators can take years of development, deployment and testing before going into operation. This CDN was deployed without impacting existing infrastructure performance. VideoExchange technology is delivered under licence paid for based on traffic, allowing KPN to extend the network as its subscriber base grows. By deploying its telco CDN, KPN was able to optimise the content flow on its network, improve quality of service as well as capacity and can now offer advanced and competitive services to content providers, helping KPN take a role in the distribution value chain. KPN's CDN has a capacity of 100 Gbps. It can support Flash, Silverlight, and H.265 streaming services. It can provide its subscribers faster Web downloads, especially for content such as sports and event broadcasts. One of KPN's first CDN customers was Eredivisie, a soccer league in the Netherlands; games are broadcast direct as PPV, a subscription service. IDATE

44 KT Since 2006, KT positioned itself in the datacenter market by deploying Internet Computing Service (ICS) in 8 different locations in South Korea. As a result, the operator has the largest datacenter in Asia (more than 100,000 m2 of dedicated space), with simulation and low energy consumption. Relying on its cloud infrastructure, KT launched its large scale offer, ICS-Ucloud CDN in 2011, which is a solution that combines cloud computing and CDN, with installations in Los Angeles, New York and Frankfurt, with other sites to come. Service pricing will be based on volume and bandwidth. Figure 37: KT's data nodes presence Source: KT KT's CDN service targets online and mobile gaming companies, content providers and domestic and international multinationals. KT particularly targets companies wanting to expand worldwide. The offer is divided among three products: downloading content, especially games and navigation systems content caching, for portal pictures and Flash objects, news sites and e-commerce media streaming, for broadcasting and e-learning Among KT's references are video game publishers such as NCSoft, direct and VOD broadcasters, as well as news portals such as ZDNet Korea. According to KT, telco CDN has some advantages over traditional CDN, based on their network knowledge, including its ability to manage large amounts of traffic by combining CDN and cloud services. KT says 90% of CDN traffic will eventually be managed by telcos. Figure 38: KT's CDN deployment Source: KT IDATE

45 KT is in favour of a federation between telco CDN and other CDN players. KT took up the question of Net Neutrality in February 2012 when Samsung tried to deny Internet access on its Connected TV for bandwidth overconsumption. The operator announced wanting to be able to establish an easily implemented rule for all platforms offering bandwidth hogging content, which, according to KT, is a danger to network stability. There could be a form of royalty paid by content providers to the operator. The operator targets connected TV manufacturers as well as OTT players such as YouTube Telstra In 2005, Telstra developed its own CDN, primarily aimed at distributing its own content across Australia. As part of its deployment of a new generation IP network, and to distribute managed and non-managed services, Telstra deployed a new CDN in 2010 by extending content distribution over a dozen "small" datacenters. The Australian operator worked with Cisco to deploy the network in nine months. Telstra relied on cloud computing, with datacenters, to control its capacity needs. Telstra plans to invest 14 million USD over the next three years to improve media content distribution, by deploying other datacenters that will be located at Ethernet aggregation points. Telstra's CDN solution is mainly used by Australian broadcasters to distribute and broadcast media content. Telstra's advantage is its strong domestic presence and its knowledge of the local media market. The operator can especially cache sports events, during retransmission, as close to the user as possible in peak periods. Its differentiation strategy is to integrate online video and content platform management solutions. Its motivation to deploy its own CDN, beyond providing an improved experience to its subscribers, was to reduce backhaul costs. Figure 39: Telstra's CDN solution Source: Telstra IDATE

46 Now Telstra is developing its offer to provide a multiscreen solution. Eventually, beyond media content distribution, the operator want to use CDN for targeted advertising based on the user's location or his preferences. Currently, the telco CDN has the advantage of accessing information such user profiles and their location, which can be used by broadcasters, content providers and advertising agencies to provide targeted services. Figure 40: Targeted advertising using CDN Source: Telstra TDF TDF is not actually a telco. It has a multi-format and multi-device terrestrial platform to manage and distribute its customers video, audio and data content to all types of devices. However, TDF acquired SmartJog in 2009, which gave it the ability to broadcast content over the Internet. TDF, using SmartJog, has been able to send audiovisual content by satellite and IP network, to about 300 television channels for international distributors. SmartJog also manages the exchange of programmes and digital rushes between movie studios and postproduction companies. Figure 41: TDF's offer (Media Services) in the value chain CDN E D I T o R S INGEST PROCESS DELIVER Source: TDF Based on its broadcast network, SmartJog is in a position to deliver programmes anywhere in the in the world from the major laboratories (Eclair, ScanLab, Deluxe, Technicolor, etc.). SmartJog also has permanent satellite capacity in DVBS2 (40 Mbps). IDATE

47 In mid-2011, SmartJog launched its new media broadcasting solution, called "MediaConnect" and dedicated to audiovisual customers, to connect devices in France. It has a network infrastructure (10 PoP in Paris and regionally) and is able to handle massive audience peaks. SmartJog expects to offer a better user experience for video consumption on all connected devices (Connected TV, PC, digital tablet, smartphone, etc.). SmartJog is able to distribute TV and radio programmes, directly or on demand, under multiple formats and to all connected devices. In mid-2012, through its SmartJog subsidiary, TDF announced the acquisition of Yacast Media, an Internet and mobile broadcaster. The offer's business model is based on sharing value with partner Internet access providers, in exchange for a guaranteed quality of service for managed flow to the end user, using the operator's network management system. With its 10 PoP, it is able to interconnect with regional ISP. According to TDF, telco CDN shouldn't disrupt its strategy. It doesn't believe in the actual sale of CDN services by CDN operators. These will do a lot of transparent caching. Implementing telco CDN resulted from the need to optimise costs as well as limit churn in its retail offer Bell Canada In 2009, Bell Canada established an exclusive partnership with Limelight to deploy its CDN infrastructure to take advantage of the infrastructure as well as the experience of a CDN specialist. The business model between the two players is based on a managed CDN model in which Bell Canada doesn't resell the infrastructure but sells a Limelight technology licence and integrates Limelight's servers into its own IP network. Limelight's datacenters are interconnected to the Bell network using a fiber backbone network. As part of this multimillion dollar contract, Limelight is responsible for operation and maintenance of the CDN network. Figure 42: Evolution of CDN business models Source: Bell Canada and Limelight Astral Media, one of the largest media companies in Canada, was one of Bell Canada's first customers to use CDN for family.ca, themovienetwork.ca and superecran.com, to distribute and monetize online content. According to Astral Media, relying on Bell Canada's CDN network for streaming video lets them avoid massive investment in servers and network infrastructure. Bell Canada's CDN can offer infrastructure to content providers to broadcast content in any format, including HD. To address any type of device, Bell's offer is based on modifying the content for the device (device detection and adaptive broadcasting). The network supports up to 3 Tbps worldwide with more than 70 datacenter around the world. IDATE

48 Figure 43: Bell Canada's CDN network Source: Bell Canada Telecom Argentina In July 2011, Telecom Argentina tested its telco CDN with a media broadcaster during the America Cup soccer tournament, by providing streaming services to radio and television stations with broadcast rights. During the test, Telecom Argentina distribution traffic at 15 Gbps for 30,000 users simultaneously, 97% of whom were located in Latin America. After the success of this test, Telecom Argentina launched a video streaming service called Arnet Play in October 2011, covering all types of content, convinced of the telco's role in local content distribution. The first PoP were deployed in Buenos Aires, Rosario and Cordoba. Telecom Argentina selected Broadpeak as equipment supplier and developed a CDN supporting between 100,000 to 500,000 simultaneous flows, for all types of streaming formats including HD content, with a capacity of more than 80 to 300 Gbps. In terms of storage, the telco has two points of presence, with capacities up to 1 PB and 25 TB network nodes. In 2012, the telco wanted to extend its network to surrounding countries. Figure 44: Telecom Argentina's CDN architecture Source: Broadpeak IDATE

49 4.2. Equipment supplier positioning Summary Massive investment by equipment suppliers Over the last two years, major equipment suppliers have been eager to position themselves in the CDN market, especially to target solutions for their operator customers. The equipment suppliers have engaged in acquisitions of CDN specialists, as well as integration and distribution partnerships. Table 3: Equipment Suppliers Cisco Acquisitions and partnerships of the major equipment suppliers Acquisition - Integration Partnership Acquisition of BNI Video in 2011 Objective Software publisher of video back-office management and analytical tools for CDN Juniper Acquisition of Ankeena CDN specialist with Media Flow Detector platform Alcatel-Lucent Acquisition of Velocix CDN specialist with Digital Media platform Alcatel-Lucent Partnership with BlueCoat Transparent caching specialist with its CacheFlow solution Ericsson Nokia Siemens Networks Nokia Siemens Networks Integration of the JetStream platform Integration of the Verivue platform Partnership with Oversi CDN Specialist with the VideoExchange platform CDN Specialist with the OneVantage platform Transparent caching specialist Huawei Joint-Venture with Symantec Development of the transparent caching solution icache Source: IDATE CDN and Transparent Caching positioning Currently, all of the equipment suppliers are generally positioned in providing solutions that combine CDN and transparent caching, based on solutions developed internally or with partnerships, as shown in the table below. Table 4: Network infrastructure Equipment supplier positioning CDN Cisco Cisco CDS None Alcatel-Lucent Alcatel-Lucent Velocix BlueCoat Transparent caching Juniper Juniper Media Flow Control Juniper Media Flow Control Ericsson Ericsson Media Delivery Network (with JetStream Video Exchange and Akamai for mobile) Ericsson Media Delivery Network Huawei Huawei icache Oversi and Symantec Nokia Siemens Nokia Siemens (with Verivue OneVantage Legend: internal solutions in blue, integrated solutions in orange Source: IDATE Nokia Siemens (with Verivue OneVantage) and Oversi With a convergent architecture, operators can cover any type of content to be distributed - managed and unmanaged - as well as maximise hosting space dedicated to CDN infrastructure and transparent caching. IDATE

50 Transparent caching techniques vary from provider to provider. Some players like BlueCoat, in partnership with Alcatel-Lucent, have based their solution on proxy servers; others like Huawei use centralised DPI architecture to "reroute" user requests; still others like PeerApp as transparent cache providers use cache servers and DPI. Equipment suppliers universally support the deployment of telco CDN, especially to help reduce CAPEX and OPEX costs required to handle traffic generated by OTT applications, as well as to generate revenue while improving the user's experience, thereby increasing ARPU with premium services. Cisco is undeniably known as the leader in the telco CDN market among infrastructure equipment suppliers, with its overall video broadcasting management offer. Heavily involved in telco CDN federation work, with a pilot completed with several operators, Cisco can take advantage of major references. Alcatel-Lucent, through its acquisition of Velocix, is also one of the leading CDN suppliers. Cisco and Alcatel-Lucent, among others, feature CDN and their video architectures in setting up their offers for their cloud and equipment customers. Figure 45: Structure of equipment suppliers CDN offers Source: Alcatel-Lucent Cisco Global but modular approach According to Cisco, the arrival of telcos in the CDN market changed the business model from B2B to B2B2C. Cisco's offer is based on supplying hardware and software, not in supplying CDN services. For Cisco, CDN is a content distribution network. It not only "caches" the content and distributes it, but also optimises the network to better deliver content to users. Cisco's CDN solution is a key element in its overall offer to manage video broadcasting, Videoscape, using the Content Delivery System-Internet Streaming (CDS-IS) platform. This platform is modular and flexible enough to meet the operator's needs. When Videoscape is managing media acquisition and eventually content conversion, the CDS-IS distribution solution acquires, streams, routes and caches the content at the edge of the network. The CDS-IS platform consists of hardware with Content Delivery Engines (CDE) on which the software, Content Delivery Applications (CDA), is integrated. This software controls content storage, ingestion, distribution, caching, customisation as well as streaming. IDATE

51 Figure 46: Cisco's Videoscape solution Source: Cisco The appliances actually can be installed on dedicated equipment or integrated directly on router blades, which reduces space and simplifies deployment. The latest generation of these appliances is the CDE 250, with three models with different characteristics (in terms of storage capacity and interfaces) to distribute content. Some models are specific for streaming TV distribution; others support acquisition and content streaming; some are dedicated to caching and streaming with storage capacities going up to 14 Tb, Cisco's solution is based on using hierarchical storage, to develop large "libraries" of content and simplify storage management. Figure 47: Cisco's CDS platform Figure 48: Ingestion, caching and distribution Source : Cisco In addition, Cisco provides managed services for appliances to control CDN architecture and do reporting in real time, required by the telcos for their internal needs (to analyse user behavior, for example) and to have accurate visibility into traffic distribution and capacity utilization. In November 2011, the equipment supplier acquired BNI Video for 99 million USD to support its Videoscape solution. BNI Video is a company specializing in reporting tools. It develops software for back-office video management and analytical tools specifically for CDN. IDATE

52 By using Cisco's solution, telcos can access various types of services, such as: content publication and management, including encoding on several formats and metadata creation; multiscreen distribution, supporting various formats and readers (tablet, connected TV, console, PC, etc.); control distribution. Extending platforms to mobile Traditionally an equipment supplier for fixed telecom networks, Cisco also supports mobile in its Videoscape solution. Actually, Cisco has integrated the packet management platform into the core of the mobile operators' networks, as a result of the acquisition of Starent Networks at the end of Cisco integrated the ASR 5000 gateway in its CDN offer, creating an intelligent and flexible routing platform. This gateway is able to distinguish between different content and different devices that originate the request and knows to make intelligent decisions on how to distribute the content requested. The ASR 5000 can actually apply specific prioritised rules to packets to smooth out traffic. It can also deliver content from the cache, like initiating conversion of a video into the most optimal format for the mobile device. Depending on the telco requests and traffic flow, the Cisco solution pushes content as far down in the network as possible in the IP equipment - down to the BRAS in the fixed network. Not all information is systematically cached. Economics justifies what is cached. In mobile networks, cache servers are installed in the packet network core (EPC in an LTE network). It optimises traffic flow and particularly video flow in high density areas. Figure 49: Cisco Mobile Videoscape Source: Cisco Considering legal discussions about transparent caching, Cisco is not sure that the expected quality can be delivered, despite the real demand that telcos have for managing OTT traffic. However, the equipment supplier recognises transparent caching as an alternative where connectivity may be limited, as in Africa for example. CDN provider recognised by cable and telecom operators The cable operators are Cisco's largest customers for its CDN offer. It supplies American players like Charter Communications and Time Warner Cable, as well as the French cable operator Numericable. But the CDS-IS platform has also been deployed by major operators in Europe, such as BT and Telecom Italia, and in Asia by Telstra. In 2010, Numericable chose Cisco to offer high quality video content from various sources for any type of device (smartphone, tablet, computer, TV). The cable operator notably deployed the CDS solution for its ability to cache and perform dynamic content conversion, surpassing traditional streaming solutions. IDATE

53 That same year, BT also chose Cisco's CDN solution to move content to the edge of its network by deploying CDS streamers. By relying on Cisco's CDS-IS solution, BT's wholesale business deployed a wholesale CDN network "Content Connect", based on a white label platform. According to Cisco, video distribution and formatting costs have been reduced 10% and overall processing time by 20%. Figure 50: CDN implementation in BT's network. Source: BT In Australia, Telstra launched a CDN project over 3-5 years for an investment of 14 billion USD. A CDN based on Cisco equipment was deployed in its Next IP network in Telstra is planning to extend content distribution to a dozen mediacenters (equipped with servers) in existing aggregation points on the 150 servers already in the network. According to Telstra, deploying its own CDN gives it flexibility not offered by CDN such as Akamai to reduce backhaul costs (content being pushed as close to the end user as possible and distributed "just in time"), while delivering a better user experience due to lower latency. CDN makes sense during peak demand, such as during a popular sports event in a specific region. Heavy involvement in the telco CDN federation Cisco is heavily involved in the work and review of the CDN federation network. One of its arguments is that the market could double in size due to the federation. The federation should be open due to the differences between operator networks. A pilot was performed by Cisco in collaboration with BT, KDDI, Orange, Telecom Italia and SFR. The object of the pilot was to analyse the economic and technical operation and to push standardisation for the federation of CDN. Cisco is also heavily involved in the IETF CDNI working group. The first phase of this pilot was completed in October 2011 and was performed in a laboratory simulating free content and multi-vendor CDN equipment in a fixed network environment. This determined technical aspects such as supporting several streaming protocols (static and progressive), dynamic content acquisition, information reporting, CDN interconnection with various architectures. IDATE

54 Figure 51: Phase 1 of the CDN operator federation pilot Source: Cisco The second phase, completed in June 2012, involved other operators such as Bouygues Telecom, Swisscom, Telstra and BICS (Belgacom's international subsidiary). This phase demonstrated how CDN federation could improve the quality of the user experience, significant for content providers, as well as it handles the growth in mobile traffic. It also included a review of the roles and responsibilities of the various players involved in the federation. Figure 52: Phase 2 of the CDN operator federation pilot Source: Cisco The third phase of the pilot will take place in the last quarter of It will include additional functionalities, integrating fixed and mobile, and cover other players in the value chain such as content providers. IDATE

55 Alcatel-Lucent CDN as business model to manage the growing amount of traffic Alcatel-Lucent's entry into the CDN market was based on the acquisition of Velocix, a CDN infrastructure provider, in The equipment supplier is counting on the synergy between its operator network experience and Velocix' experience in CDN. It wanted the shortest timeto-market by taking advantage of Velocix's knowhow. The equipment supplier estimated that the transfer of knowledge will take one or two years. According to Alcatel-Lucent, due to the on-net CDN 3, operators are able to cache and distribute content as close to the user as possible, and reduce video transport costs (video being the most important for the equipment supplier), as well as increase revenue (by monetizing the user experience). Launched in 2010, Alcatel-Lucent's offer is a key element in its High Leverage Network (HLN) multi-access IP network for distributing data and Internet content. Based on the Velocix platform, its CDN offer was developed specifically for operators deploying their own CDN networks. Alcatel-Lucent's solution, Velocix Digital Media Platform, includes four types of software applications. The application can be combined depending on the operators' needs. Velocix Delivery is an appliance to distribute requested content to the device that made the request. The cache location is determined based on transport savings and analysing the popularity of the content. Once the location is determined, the Velocix Delivery application distributes the content at a set speed. Velocix Publishing, dedicated to content ingestion, is used to analyse content and clip it into metadata before storing and redistribution. This appliance is notably used by operators of wholesale CDN. The platform also supports video streaming content distribution using the Adaptive Bit Rate (ABR) protocol. Velocix Storage is used for storing content and acts as a "depot" (with solutions with flexible storage capacity, from 4.5 Tb to 137 Tb which can be extended up to 1,100 Tb per appliance. Velocix Service Node is the appliance including operational services such as interface management, usage reporting, and network performance monitoring and billing applications. This application is particularly important for operators to understand user consumption and thus personalise their offers based on the reporting. Figure 53: Velocix Digital Media Delivery platform Architecture Source: Alcatel-Lucent In addition, Alcatel-Lucent's offer includes Velocix Managed Service, an optional solution that includes remote CDN network management by Alcatel-Lucent, leaving operators to concentrate on opportunities from their CDN. Velocix Managed Service is operated by teams from the original Velocix. 3 on-net CDN differentiates it from traditional CDN providers considered as off-net CDN. IDATE

56 Alcatel-Lucent was one of the first to create a partnership to develop multiscreen solutions, including several players including BlueCoat, Microsoft, theplatform, etc. Its Velocix Digital Media Delivery platform supports multiscreen content distribution (TV, tablet, computer, mobile phone, etc.), providing different user experiences. Actually, Alcatel- Lucent established a partnership with theplatform in June theplatform specializes in publishing and managing online video for multiscreen distribution. Alcatel-Lucent deployed a common video platform with advanced video management functionality and optimised broadcasting using CDN, accessible on all types of devices with high security and flexibility. For Alcatel-Lucent, CDN makes sense at aggregation points (BRAS) in the current topology, and further down than the edge of the access network on fixed infrastructure. However, in the mobile network, there is no interest in implementing CDN at base stations. Figure 54: Integration of CDN in the existing network Source: Alcatel-Lucent Partnership with BlueCoat for transparent caching The Velocix Digital Media Delivery platform doesn't natively support transparent caching. Not wanting to develop its own solution internally, Alcatel-Lucent signed a distribution agreement in September 2011 with BlueCoat for its Web content caching and distribution appliance, CacheFlow. This solution can be sold in addition to the CDN platform. By deploying CacheFlow, bandwidth consumption - especially OTT content - can be reduced by caching the content transparently, generating savings in bandwidth for operators and improving user experience. For a telco, the Velocix platform and the BlueCoat application are complementary to distributing managed and unmanaged content. However, according to the equipment supplier, obstacles to development of transparent caching today are based on the development of HTTPS. Transparent caching is essentially useful in emerging countries to maintain a certain level of user experience. Customers Alcatel-Lucent primarily provides its CDN solutions to cable operators in the United States as well as telcos like Verizon. Among them are the British broadband provider Talk Talk and Orcon, the New Zealand provider, to offer wholesale CDN. Actually, Talk Talk selected Alcatel-Lucent's CDN solution to build its own distribution network over a period of three years. The provider sends British video content over a secure IP network for a better user experience and reduces its peering costs for Internet connectivity. Alcatel-Lucent encourages the worldwide federation of CDN networks. From a technical point of view, the equipment supplier estimates that there are no obstacles even on the issue of platform compatibility. A CDN network based on a wholesale model would make most sense. However, the equipment supplier is not as heavily involved in the work on federation as Cisco. IDATE

57 Juniper Networks According to Juniper, operators have an undeniable interest in positioning themselves in CDN because they control both the network and users. CDN helps operators deploy new services and monetize them on the network. Beyond providing CDN solutions for operators, the aim of Juniper is to leverage on its expertise and leadership in IP service provider networking. Technically the idea is to be able to manage the network and content delivery infrastructure on a single platform by combining media infrastructure solutions with Juniper s high-performance networking platforms Juniper s general objective is to provide Media Flow, its CDN solution that reduces costs through equipment consolidation, lowering server hardware CAPEX and OPEX by 70% and delivering 10 times more media than traditional caches. Indeed, according to Juniper, its solution provides up to a 10-to-1 reduction in the number of servers needed to deliver the same amount of media, with a similar reduction in transit network and media delivery costs. To achieve its objectives, Juniper proceeded with several key acquisitions with Ankeena, a CDN provider for the Media Flow Director platform (April 2010) and Bitgravity for the integration of CDN Service Management Layer including network monitoring and reporting (January 2012) The Media Flow Solution is based on the key components that enable service providers to build an efficient CDN while leveraging Juniper s Junos router application. Actually, Media Flow Solution provides Internet streaming video and other rich media content on a converged, managed, and efficient architecture. The solution combines storage organisation, multitier caching, and network optimisation with storage systems: Media Flow Controller (MFC), caching software that provides awareness and analyse content. MFC places objects into storage based on type, size and popularity and dynamically manages content based on demand. Media Flow Publisher, which supports adaptive streaming formats and allows multiscreen video services delivery. VXA Series Media Flow Engine (caching hardware platforms) to optimise Media Flow Controller and support up to 7 TB of cache with a capacity ranging from 1Gbps to 20Gbps.. Figure 55: Juniper's MediaFlow Controller Source: Juniper Networks In the CDN field, Juniper also works with many partners: Triveni Digital specialised in video monitoring and analysis application Cedexis providing cloud based solutions Openwave Systems to integrate the Media Optimizer, a mobile video optimisation tool, into Juniper s Media Flow IDATE

58 SkyTide for its Insights CDN analytics and reporting solution Verimatrix for an Internet TV content security platform. The placement of Media Flow Solution components depends on network size and topology, as well as on the end goals desired by the operator. Figure 56: Media Flow Solution position within the network Source: Juniper Networks Edge: close to subscribers. This placement minimizes online transit traffic across the backbone, expediting content delivery to end users. When deployed at the network edge as a transparent cache, Media Flow platform addresses network optimisation by caching and delivering popular content from a local cache device thereby reducing the bandwidth, protecting the resources, and subsequently driving down the costs. Figure 57: Media Flow at the network edge Source: Juniper Networks Mid-tier: caching server is placed near high traffic exchange points. This placement offloads session termination traffic from the origin server and increases bandwidth. IDATE

59 Origin: content and cache are collocated to maximise efficiency of video distribution in the backbone. By caching at the origin, repeat requests need only hit the locally cached content and thus, do not need to reach the origin server. As such, content delivery is accelerated, as is the ability to scale the performance of network storage file servers and Web servers. (CAPEX reduced by server consolidation). Juniper also delivers optimisation solutions in order to deliver video on mobile networks. Here the MediaFlow platform is associated with the Openwave Media Optimizer and Openwave Traffic Controller. This solution helps particularly to reduce traffic across the RAN. Figure 58: Juniper's solution for mobile network Source: Juniper Networks Huawei CDN is part of Huawei's global strategy to provide solutions to optimise content distribution, especially video. Actually, video has a heavy impact on telco distribution costs. Huawei helps telcos deploy their own CDN to reduce backhaul traffic congestion, especially on the mobile network. Also, CDN gives access to new sources of revenue to monetize their networks. According to Huawei, there is strong potential for expanding telco CDN, especially in Asia, due to the fast adoption of streaming. Huawei developed its own CDN offer under the name of SmartCDN. According to the equipment supplier, this platform implements new business models to improve the user experience and help operators find a position in the media value chain. SmartCDN supports several services on a single system and dynamically adjusts resource allocation, making it easier for operators to make modifications required to meet service needs. Dedicated to deployment on the network edge, the SmartCDN platform revolves around "small boxes" to distribute video, including: Cache servers, to cache and distribute popular content, for conversion to adjust formats based on bandwidth with storage capacity from 40 SATA cards of 500 Gb to 4 flash memory cards of 3 Tb. Media control servers, which perform routing (managing 20,000 HTTP requests per second) and which can also support other protocols. Storage servers, with a maximum capacity of 30 PB (for 100 clusters each with 14 servers of 22 TB), Content management servers, dedicated to ingestion and pre-distribution of content as well as for collecting network and connection data so telcos can determine statistics on network conditions and user preferences. This data analysis is used to provide value added services. IDATE

60 Network management servers, for operations and maintenance as well as monitoring key indicators (KPI) and network status. Figure 59: Huawei's SmartCDN topology Source: Huawei However, according to Huawei, one of the drawbacks of CDN is the ability of operators to come to agreement with content providers for access to content and its distribution. This is why Huawei is also positioned in transparent caching, as part of its partnership with Symantec to develop the icache solution. This alliance gives Huawei access to Symantic's expertise and product portfolio in storage and network security. icache by Huawei-Symantec, based on DPI, caches and accelerates content distribution. icache can technically determine if the content should be cached or not, by fragmenting the content, and also tracks users and analyses network traffic. Beyond improving the user experience, the main objective for icache is to reduce pressure on bandwidth and lower the associated costs (estimated at up to 50%). Tests have shown reductions of up to a third in bandwidth using the icache solution. Figure 60: Location of icache in the network Source: Huawei IDATE

61 Figure 61: icache architecture, Huawei-Symantec transparent caching solution Source: Huawei Huawei favours a federation of CDN networks, but has not participated in the tests. According to Huawei, the federation should be formed even if there is no standardisation yet Ericsson CDN: for traffic management Ericsson's interest in CDN is to provide premium content distribution and ensure end-to-end continuity. Actually, the equipment supplier is essentially involved in e-commerce, financial and B2B applications. However, Ericsson encourages CDN deployment by operators to improve user's TV experience by CDN solutions, transparent caching and traffic smoothing, to reduce bandwidth costs and generate new revenue. The CDN research group investigates content distribution technologies including traffic redirection, caching, eliminating redundancy, and content migration. As with all equipment suppliers, Ericsson used its network expertise and understanding of telecom operators as well as its international implementation (local presence and expertise) to position itself in CDN. Ericsson established a key partnership with JetStream to integrate its CDN solution, VideoExchange, in the core of its Media Delivery Network (MDN) platform. Ericsson can offer a video distribution solution to operator that is based on a combination of a CDN infrastructure and its IPTV products. As part of this partnership, JetStream is responsible for training Ericcson personnel on the CDN platform so that the equipment supplier can directly help its customers for design, architecture and integration of VideoExchange. Other partnerships are also in place, with Edgeware to provide video equipment (especially VOD servers), and SkyTide for analysis tools and reporting. Overall, the platform is designed to view personalised data from anywhere, anytime, on any device, while letting operators efficiently transport the surplus traffic generated on their networks. The network supports video on demand and OTT video content services, using the same infrastructure. Ericsson's Media Delivery Network platform, based on JetStream's CDN solution, includes three main functions: Content management, with broadcasting and content distribution integrating classic CDN functionalities: ingestion, storage of content metadata, caching, streaming, conversion and other control functions. Transparent Internet caching, where the key element is DPI, with content management functions to validate that the stored content is properly synchronized with the original, making the link between the URL and the content itself. Units to improve services and performance using cache servers and a content management system (CMS) as well as compression and optimisation to modifying content for the mobile network. Targeted advertising can be on any platform for any screen. IDATE

62 These functions could be deployed in both fixed and mobile infrastructures. Figure 62: Ericsson's Media Delivery Network functionalities Source: Ericsson Ericsson supplies its CDN solution to Tata Sky, the direct television operator in India. The network was deployed in several PoP. According to Tata Sky, the choice was based on the Ericsson and JetStream combination for pushing content to the edge, because of its limited backhaul capacity, rather than use traditional caching. Being the leader in mobile infrastructure, Ericsson is more interested in mobile distribution, which is very different from fixed. This is why Ericsson and Akamai formed a strategic partnership in Beyond their respective expertise, the two companies can also take advantage of strong rapport with CDN players: Ericsson with mobile operators and Akamai with content providers. The objective of the combination of their solutions is to provide cloud accelerator services within the mobile network ("Mobile Cloud Accelerator"), including expertise in mobile operator infrastructure, CDN and RAN priority management, for an endto-end distribution solution. The Indonesian operator, Telkomsei, chose this common solution to deliver a better customer experience. The operator stated that download time was reduced by more than 70% using Mobil Cloud Accelerator. Ericsson supports a federation of CDN operators, but recognizes the lack of standardisation. Not directly involved in the CDN federation, Ericsson monitors the work on defining a common standard through international organizations such as ETSI, IETF and ATIS. According to the equipment supplier, a global standard would lead to content providers accepting this new model, which would change the CDN market. IDATE

63 Nokia Siemens Networks CDN for more intelligence in the network Nokia Siemens Networks is also one of the equipment suppliers that are investing in CDN with R&D efforts, especially to bring more intelligence to the network and manage OTT traffic. For the equipment supplier, operators really want to deploy their own CDN infrastructures to make their networks more efficient and to monetize their content distribution capacities while improving the user experience. Nokia Siemens Networks decided to use partnerships with Verivue and Oversi to provide CDN and transparent caching to operators. It uses Verivue's platform, OneVantage CDN, to integrate content distribution solutions. In October 2010, Nokia Siemens Network and Verivue signed a distribution agreement. The Verivue platform is a software suite including caching, content distribution, routing, origin server, transparent caching and analysis and platform management tools. Nokia Siemens works mostly with cable operators like UPC as well as with European mobile operators. Nokia Siemens offers several solutions to mobile operators to manage the explosion in traffic, most notably its Evolved Packet Core (EPC), a constantly changing solution at the core of the mobile network managing broadband connections. The equipment supplier provides solutions for modifying content, optimising distribution, policy control, offloading on Wi-Fi, caching and content security. In particular, Nokia Siemens delivers monitoring solutions and QoS differentiation based on DPI technology, an element integrated into the Policy Control Resource Function (PCRF) at the GGSN gateway interconnection between the mobile packet networks and the data networks HP One of the largest manufacturers of servers, HP is positioned in cloud computing and has extended its business by now also being present in CDN through a cloud computing approach. HP is heavily involved in providing content distribution solutions to operators (HP's annual revenue from operators is 7 to 8 million USD). According to the equipment supplier, operators need to be aggressive in content distribution to monetize their networks. Source: HP Presented to the public in beta in June 2012, HP officially launched its HP Cloud CDN service, as well as its public cloud service, HP Cloud Storage in August For its object cloud storage service, HP is targeting Web content storage, data backup and archiving, post-attack recovery, as well as the "Big Data" market. IDATE

64 For its CDN offer, HP decided to use Akamai's platform, giving it access to a CDN network with a global presence (more than 80 countries) and significant size (more than 100,000 servers in more than 1,900 installations), for distributing all types of content. For HP, the partnership with Akamai lets it offer a highly flexible solution, ensuring minimal latency to access objects through caching and network optimisation services. In addition, HP Cloud CDN runs on HP Storage Object Cloud to store copies of content, making it easier to access data. Actually, the two solutions are independent but customers can choose HP Cloud Object Storage and rely on HP Cloud CDN to deliver their data if needed. CDN guarantees high transfer rates so companies can deliver large objects to end users. HP Cloud CDN includes: caching, where content is cached and distributed through the network of servers closest to the user; progressive downloading HTTP compression, to compress large objects and lighten bandwidth load; and SSL encryption, for more security during distribution. HP Cloud CDN billing is based on the volume of data transferred, and depends on the customer's billing location. Table 5: Pricing based on customer location Customer location North America, Europe and Latin America Japan, Hong Kong and Singapore $0.19 rest of world $0.39 Source: HP Price for the first 10 TB $0.16 per GB HP's telco CDN offer was released in July 2012 under the name of SpeedVideo CDN. It s a solution that integrates content production and distribution using cloud computing capabilities. This solution includes optimising the various elements of CDN by relying on network peering, content "fragmenting" (large objects as well as video), cloud computing, adaptive streaming to adapt to the various Internet formats, while integrating into the telcos' network topology. SpeedVideo CDN also includes additional functionality such as transparent caching, coding/conversion, and DRM. For architecture, HP's approach is both global and local: the CDN core with central nodes for content distribution via P2P; distributed notes for content storage and distribution to end users. Through this offer, HP can target any type of telco, broadcaster, content provider as well as e-commerce sites. IDATE

65 Figure 63: HP's SpeedVideo CDN global architecture Source: HP 4.3. CDN positioning Summary CDN players have already revised their strategy and did not wait for the arrival of telco CDN to adapt. As shown in the first section, CDN players have diversified so that they are less dependent on only selling bandwidth and competing against low-cost offers such as Amazon's. They initially specialised in the sale of value added services (traffic analysis, marketing services). These services used to make up a significant share of revenue (50% at Akamai and 30% at Limelight). Based on their distributed server architecture, CDN players have also turned to cloud computing to offer Infrastructure as a Service (IaaS) services. Competition from telco CDN With the arrival of telcos where they expect strong competition, their market share could drop considerably. The traditional CDN market is starting to consolidate: acquisitions between them (Akamai and Cotendo); acquisitions of certain players by telcos themselves (KDDI-CDNetworks). Actually, for them, the deployment of telco CDN is potentially a source of competition, because they still depend on existing business models (such as dedicated sections). Telco CDN could make traditional CDN obsolete, caching being done lower in the network. This doesn't necessarily guarantee better performance (especially if the caches are still too high or too low in the network. The congestion problems are found mostly in the middle-mile (i.e. in the backbone and peering points). But they should provide at least equivalent performance. The quality should also be better in direct interconnections of a content provider or a CDN transporting a content provider's traffic. CDN deployment by ISP is more susceptible to drive a loss of traffic and margins (and revenue) for traditional CDN. They could actually be outdone by content providers (knowing that the biggest ones are doing it already, but not when they really need QoS). The can also interconnect directly with the ISP, which could cause new costs, albeit low if the ISP is targeting optimisation of its network costs. The changes might not be as bad as it currently looks to the traditional CDN players, especially when it comes to CDN-telco CDN IDATE

66 interconnection. QoS could significantly improve, attracting more traffic for the two entities to manage due to the increase in usage and revenues from content providers (benefiting from the increase in QoS). In addition, using this solution, traditional CDN players could potentially decrease costs, thereby maintaining their margins. It could be possible that the entire CDN market would grow, and that the overall impact for CDN players would end up being a wash. Trend toward licences and managed services The biggest CDN players, however, are keeping an eye on these developments (though they contest the technical aspects of it, assuming that the operators don't know how to make a true CDN), even if they could eventually benefit from it. The most probable is that their situation gets worse because the telco CDN are new competitors, at best new intermediaries in the value chain, with whom they would have to share revenue. To face this loss of control, other than revenue from value added services that the operators aren't able to quickly recreate, some CDN players are looking to position themselves as technology providers or services for ISP wanting to position themselves in CDN, through distribution partnerships and white label sales of technology, in direct competition with equipment suppliers. As a result, they have decided to licence their technology to help the telco's deployments. For example, Deutsche Telekom chose to partner with Edgecast and AT&T chose Edgecast and Cotendo to accelerate the time-to-market of its own CDN product (some rumors say that AT&T could opt to resell the licence of a major CDN player (Akamai or Limelight). Most of the players now offer such a play, entering into direct competition with equipment suppliers. Regarding the CDN federation (or interconnection of CDN), even though they were (very) reluctant at the beginning, they have changed their strategies in parallel since the telcos have implemented or threatened paid peering to borrow their networks (traditional CDN would then have to pay (by Mbps) the telco (price generally less than the price of transit). Table 6: Positioning of various traditional CDN around telco CDN Player Telco CDN initiative Geographic area Customers (nonexhaustive list) Akamai Limelight Networks CDNetworks Dedicated product (06/2011) Dedicated product (02/2012) Acquired by KDDI, no dedicated product America, Europe Essentially Americas Bell Canada, Bestel Edgecast Dedicated product Americas, Europe T-Mobile, AT&T (partnership in 2011) Bitgravity Chinacache JetStream Acquired by Tata, no dedicated product No dedicated product, as such Dedicated product for several years Asia Asia China Northern Europe Highwinds No dedicated product Americas Amazon (see content players) Source: IDATE No dedicated product Americas, Europe Shanghai Telecom KPN, Telenet, Mediakabel, Telenor IDATE

67 Akamai Akamai is the worldwide leader in CDN services. Its markets are the Americas and Northern Europe. In 2010, Akamai made billion USD in sales. In 2012, it reached billion USD, a 12% increase. The CDN business was developed through a series of acquisitions, such as Cotendo in early Akamai's strategy is heavily weighted towards providing value added services, due to the fierce competition in distribution and sale of bandwidth markets. To follow this strategy, Akamai increased its acquisitions of companies specialising in marketing and audience measurement tools. Media customers are no longer the source of revenue growth rate. Corporate and e-commerce sites are the only ones generating double digit growth, using cloud services offered by Akamai. Though Akamai's expertise is acknowledged on the fixed Internet, it is much less so on the mobile Internet. On fixed, Akamai's CDN offer is based on its imposing inventory of cache servers distributed around the world. On the mobile network, Akamai does not offer caching services, strictly speaking. Actually, it reuses servers from the fixed network, accessible by GGSN (mobile gateway). The basis for the purely mobile offers relies essentially on adapting content to the device (distributing content in the proper format), in proper encoding (using adaptive encoding technology), and on caching services on the fixed network. In addition, Akamai only has servers, not a telecom network as such (Limelight owns a part of the network infrastructure). Telco CDN business Although initially skeptical, as were most of the CDN players, Akamai announced the release of a product based on reselling technology targeting telcos in June This offer is called "AURA" and is based on four distinct products for operators: offer based on traffic acceleration (installation of Akamai servers in the telco's own network); CDN managed services (for the operator's own content); technology licensing (purchase of software tools by the operators); offer based on federation (allowing the operator to distribute content outside of its own network, even around the world). IDATE

68 Figure 64: Description of Akamai's portfolio of services Source: Akamai Figure 65: Illustration of telco CDN deployment based on Akamai technology Source: Akamai Figure 66: List of Akamai's offer for operators Source: Akamai IDATE

69 CDN federation Akamai has refuted the idea of a federation or interconnection between CDN (traditional CDN and telco CDN) for a long time. But since 2012, Akamai has been extolling the virtues of a CDN federation to increase coverage as well as improve user experience. Figure 67: Extract of a presentation on CDN federation, May 2012 Source: Akamai Mobile CDN During the MWC 2011, Akamai and Ericsson announced a strategic partnership to distribute content over the mobile network. This solution doesn't seem to address video content especially but services with substantial margins able to pay for this type of technology. Figure 68: Types of applications targeted by the Ericsson-Akamai partnership Source: Ericsson-Akamai As part of this agreement, Akamai wants to rely on Ericsson's expertise, which manages billions of subscribers through its outsourcing contracts with operators around the planet. The objective of the partnership will be to prioritise traffic. IDATE

70 Figure 69: Outline of the Ericsson-Akamai partnership Source: Ericsson Nevertheless, in the near future (no specified timeline), Akamai expects to install equipment (or more precisely software solutions) very close to the BTS, at the foot of the cell towers. Figure 70: Akamai's view on mobile: extend its presence to the foot of the tower Source: Akamai IDATE

71 Limelight Networks Introduction Limelight Networks is a CDN player based in the United States. It is considered to be the second place CDN player in the world in terms of sales, but its primary market is restricted to North America, unlike Akamai, which has a strong presence in Europe (in addition to North America). Though its revenue is still growing, Limelight has not been profitable (except in 2009), which regularly causes speculation on its possible takeover. Figure 71: Growth in Limelight Network's revenue and net profit $ $ $ $ Revenues Net (loss) income $- $(50 000) $( ) Source: Limelight Networks Value added services represent about 30% of total revenue. Limelight, like its main competitor Akamai, has increased its acquisitions of companies specialising in marketing and on-line advertising, like EyeWonder and Kiptronic. Figure 72: Value added services' share in Limelight revenue Source: Limelight The other element in Limelight's diversification strategy is cloud computing. This also is not a unique strategy (all the major CDN are doing it to). IDATE

72 Telco CDN business At the beginning of February 2012, Limelight Networks unveiled its offer to telcos, "Limelight Deploy", to provide a technology licence, just like that proposed by its direct competitors (Akamai and Edgecast in the lead). Limelight Networks was the last CDN provider (among the major players) to convert to these types of offers. Limelight Networks has worked mostly for Bell Canada since 2009, as well as the Mexican operator Bestel since mid Figure 73: Limelight Deploy description Source: Limelight Networks Edgecast The American company, Edgecast, established in 2006, now has a network of 21 PoP on four continents, including 5 in Europe, 4 in Asia, 1 in Australia and the rest in the United States. These locations were strategically selected to be as close to the Internet exchange points, such as Amsterdam. These PoP have enormous computing and cache capacity and are directly connected to the major backbones. In 2011, Edgecast was listed as having the third largest CDN network, had doubled its sales and announced being profitable since Initially, Edgecast decided to focus mainly on small objects such as e-commerce sites, but now its CDN also supports large objects like video. Since 2010, the network also supports innovative video technologies like Adobe's Flash Media Service and Microsoft's Smooth Streaming. Nearly 2,500 billion objects have been delivered by Edgecast's network since early 2012, more than 2 billion objects delivered per hour in January 2012 compared to 1.5 billion in January 2011, when the network transported 4% of worldwide Internet traffic. IDATE

73 Figure 74: Edgecast's Network Source: Edgecast In September 2011, Edgecast signed a partnership with PeerApp to provide an offer combining CDN licence for telco CDN deployment with PeerApp's "Ultraband" transparent caching platform to manage distribution of OTT content. Operators can rely on this combination to distribute managed or unmanaged content, while reducing network costs. Actually, the combined architecture helps operators implement a unique package for configuring content caches, without agreements with content providers. In addition, combined solutions include performance monitoring and platform administration through a single interface. Figure 75: Edgecast and PeerApp partnership proposition Source: Edgecast Announced in January 2012, Edgecast also released its CDN licence offer. This announcement followed years of collaboration with a major American operator, as well as a licence agreement with PacNet, a major ISP in Asia Pacific. As part of deploying PacNet's CDN network using Edgecast technology, the POP, located in Japan, took up an important role and is used as a gateway between Asia and North America. Before, Edgecast most often positioned itself outside the United States as a partner to ISP (either reseller or white label) that did not want to develop their own CDN but preferred to rely on a third party solution: this is what happened with Deutsche Telekom, Telus and AAPT. The reseller business model is based on payment of a royalty (tax on a licence). IDATE

74 Today, beyond reselling CDN services, the company works with telcos to manage their CDN networks - especially with two Asian operators - and to provide a solution that they can use to deploy their own CDN, and even for CDN federation. In this case, the business model is based on traffic (per Mbps) and by delivered volume (Gb). Figure 76: Edgecast's Vision Source: Edgecast Edgecast encourages the federation of CDN operators. By proposing a CDN licence offer to operators, the company prefers the development of federated CDN networks. Edgecast supports the federation of telco CDN so that operators can take advantage of more capacity by combining the capacities of various operators, as shown in the diagram below. In addition, operators can target new markets and generate additional revenue. According to Edgecast, traffic and monetary exchanges between operators already takes place, which removes potential obstacles in this configuration. Complications could arise for federated telco CDN in pricing, SLA and network monitoring as a result of expectations for different levels of quality in each market. Figure 77: Telco CDN Federation Source: Edgecast IDATE

75 JetStream JetStream, a European company initially specialising in streaming video, has positioned itself as both a traditional CDN (covering Europe) and provider of technologies to operators to develop their own CDN. CDN Since 1996, the company has developed key elements in CDN with tools for replication, redirection and reporting sold under the name of Streamzilla. For JetStream, intelligence is not in the equipment, so the company provides an agnostic solution, independent of network equipment suppliers. JetStream has no network, but has servers. The company relies on 20 PoP in Europe, which makes it a major player in Europe, behind Akamai. According to JetStream, other CDN providers only have 3 or 4 PoP in Europe. In particular, the company has access to more than 100 Gbps through AMS-IX, for a total of 750 Gbps of connectivity, and relies on about thirty transit providers such as Tata, TeliaSonera, Level3 and Interoute. JetStream clearly has a particular interest in premium streaming content (notably sports), targeting hosts and content providers looking for Web acceleration. The company specialises in video-on-demand distribution for live video and audio where traffic is highest. According to JetStream, its solution is profitable because no money is spent on hardware and the return on investment is shorter, for better performance. The company has no ambition to become a global player but wants to provide quality through SLA s that guarantee streaming services. Figure 78: JetStream's Streamzilla CDN network Source: Streamzilla Telco CDN JetStream also provides its CDN technology to operators under licence. Installed in existing equipment, its CDN solution for telcos, based on VideoExchange technology (VDO-X), is a content distribution software that doesn't use cache technology - originally not designed to support large objects such as video. VideoExchange includes elements of CDN in common with Streamzilla, especially traffic offloading characteristics, creating a sort of global CDN. According to performance studied, CDN can be deployed on the edge or at the metro level of networks for telcos, while it is more effective at the head of the network for cable operators. JetStream has more than 500 customers and has targeted ISP and operators since Several operators rely on VideoExchange including KPN, Telenet, Mediakabel, Telenor and content owners such as Endemol. JetStream's business model is based on a licence with a monthly royalty and on traffic, under a three year contract. Resources can also be bought by paying by consumption per TB or by subscription (generally per Gbps). JetStream says it can save up to 50% for telcos using VideoExchange. IDATE

76 JetStream has several partnerships, notably with Ericsson, who uses VideoExchange as the basis of its CDN Media Delivery Network offer. For its multiscreen strategy, JetStream expanded its partnership program in 2010 with established players like Edgeware for CDN appliances, IBM and Ericsson for system integration, Adobe, SkyTide (analysis and reporting tools), PeerApp for transparent caching, etc. In 2012, JetStream established a partnership with Oversi on transparent caching, resulting in a unique architecture for delivering any type of content by combining CDN and transparent caching. As for mobile networks, JetStream has optimised its technology, but it is still hard for CDN to identify the user's location and efficiently deliver content. JetStream has established a partnership with Saguna Networks, an innovative provider for speeding up services and content distribution at the radio access level, notably on HSPA and LTE networks. Finally, according to JetStream, the problems with CDN federation were brought up as early as 2002 by Dutch operators, but there still is a fear of loss of control of the network as well as a loss of quality. In addition, the company suggests that there is no viable business model under a federation of CDN Transit provider positioning Summary Traditional transit providers are developing CDN in addition to their transit business, using acquisitions and partnerships (Level3-Savvis, Tata-Bitgravity, etc.). This way, they can offer a complete portfolio of Internet traffic distribution services. The "commoditisation" of the transit market, with extremely low rates, is pushing transit providers to reposition themselves in other complementary markets, including CDN. The CDN business is actually naturally complementary to the transit business, already having connectivity (large capacity) and datacenters around the world. By adding technology (such as cache servers), transit providers can have a very competitive offer compared to traditional CDN (who are also their customers), in particular for video distribution and the associated volume. Because of their economic situation (most of the players operated at a loss), all of the players have an interest in migrating towards a CDN model in addition to their transit business. This is particularly true for the largest ones, because of their wider reach. The arrival of telco CDN doesn't directly change the game. Players that had a true need for QoS already went through a CDN (see direct peering). They will continue to use the same approach, in addition to telco CDN. Some Web/content service providers could contend with transit up to the access network, and then pay a telco CDN for access. However, it's still not clear that it would meet the quality required, being only guaranteed by the last mile. Such a development would encourage transit providers to turn to CDN, with potentially a CDN interconnection. Table7: Various positions of major transit providers Players CDN Business Local geographic area 4 Details Level3-Global Crossing Tata Acquisition of Savvis in 2007, CDN offer Acquisition of Bitgravity, CDN offer North America Europe Europe, North America, but especially in Asia (India) Acquisition of Global Crossing Strong partnership with Equinix 4 Even though by definition they have a global footprint (operations/undersea cable management), most have a specific geographic business. IDATE

77 Players CDN Business Local geographic area 4 Details Cogent None North America Europe NTT-Verio None Asia Source: IDATE Transit business (IP transit only) No strategy for higher level services Level3 Level3 Communications is a network operator with a global presence including the United States, South America, Western Europe as well as certain cities in Asia. Considered as a pure transit operator, Level3 got into CDN in 2006, with the acquisition of Savvis' CDN shares. Netflix and Apple are some of its most important customers. The contract with Netflix caused some ISP to retaliate against Level3, whose traffic exploded. Litigation with Comcast, the cable operator, had bigger repercussions. Figure 79: Description of Level3's services Source: Level3 Level3 has built a true foundation compared to those offered by traditional CDN. Actually, it didn't stop at caching services (storage in cache servers). It also has distribution and billing services. Figure 80: Description of Level3's "Intelligent Traffic Manager" Source: Level3 IDATE

78 Regarding network interconnection, Level3 recently modified and updated its peering policy by adding the bitmile concept. This concept is based on the idea that, for Level3, the cost of transporting a bit (or an octet) also depends on the distance the bit goes, while the asymmetric ratio only compares the bits sent to bits received. According to Level3, the bitmile concept more accurately reflects the true cost of transport. Nevertheless, this criteria is very complicated to meet and implement. As far as financials, even though sales have increased, Level3 has not been profitable for several years, and profits have dropped since Figure 81: Change in various financial indicators (Million USD) Revenue Operating income Source: Level3 Level3 also has a department (Vyvx) dedicated to broadcasting video content (sports and news) over backhaul. In April 2011, Level3 acquired its competitor, Global Crossing. The combined share of Level3 and Global Crossing ("Level Crossing" on the diagram) reached 55% of traffic worldwide. The combination of the two companies provides 160,000 km of fibre and infrastructure in more than 45 countries. Figure 82: Key metrics for the Level3-Global Crossing combination Source: Level3 IDATE

79 Figure 83: Effect of Level3's acquisition of Global Crossing Before acquisition After acquisition Source: Renesys Blog The diagram below shows the combined network of the two operators Figure 84: Combined networks of Level3 and Global Crossing Source: Level Tata Tata Communications is a network operator based in India. It has a huge undersea cable network as well as a "Tier 1" IP network. It also leases datacenters as well as collocation services. Tata has also made several acquisitions like Teleglobe (Canadian company) and DishnetDSL (Indian operator). In 2011, Tata recorded sales of 2.5 billion USD but at a loss (-182 million USD). IDATE

80 Figure 85: Tata Communications' product line Source: TATA Its CDN offer is provided by Bitgravity (after its acquisition in 2011), a traditional CDN player. This CDN is very focused on video distribution as well as video games. The CDN architecture includes dedicated nodes in the United States, Europe and Asia (especially in India), with several PoP in these areas. IDATE

81 Figure 86: Detailed description of Tata's CDN offer in India Source: Tata Cogent Cogent is an American player specialising in providing transit services. Unlike other transit operators, Cogent has no CDN services. It is considered to be a "simple pipe". It also does not offer value added services because it only manages the first three layers of the network (no DPI technology). It's most important customer helps it achieve 5% profit. Cogent is also known for being the connectivity provider for the Mega galaxy (Megavideo, Megaupload). Figure 87: Change in Cogent's revenue (Million USD) Revenues Operating income Source: Cogent IDATE

82 4.5. Major content provider positioning Summary Internet leaders own the Internet infrastructure Though most major Internet players directly manage their infrastructure (servers, datacenters, hosting, etc.) and their AS, only a few of them have a major role in distributing their traffic. The major service providers, in particular video content providers, are induced to build their own transport network. They rely on: lower and lower transit rates; implementation of proprietary infrastructure like CDN, by installing their own servers at the edge of the access networks. As a result they end up being virtual operators of a transport network, combining transit contracts, their own infrastructure and server inventory. Only the major service providers have the resources to amortize their own investments in this approach: Google/YouTube, Yahoo, Microsoft, Amazon, ebay, Dailymotion and Facebook, and more recently Netflix. Google was the first Internet player to massively invest in Internet infrastructure (investing in all levels, including services and energy sources for datacenters). Since then, several major players have done the same, like Yahoo! and Microsoft, and even Amazon. By dispersing cache servers in various datacenters, the players are able to create their own CDN with at least partial coverage. All the major Internet players in terms of traffic, i.e. the Internet giants and the major video providers, could eventually adopt a similar approach. Major media players in multi-cdn mode Though the major Internet players are developing their own infrastructure, the major media players, used to paying for distributing their content over other means, are expected to rely on essentially traditional CDN (and would eventually be interested in telco CDN). They feel that the CDN market is a commodity (not requiring them to invest) and that the QoS from CDN is adequate. The largest players offering premium content (VOD, catch-up TV), like ABE and Hulu, generally take a multi-cdn approach, relying on three or four providers at the same time. This keeps the providers in constant competition, and lets them adapt to performance in real time and use the most efficient CDN at any moment. The often manage storage and DRM problems themselves. They rely most often on internal tools as well as load balancing solutions such as that of Cedexis. Segmentation of approaches based on size and QoS requirements The smaller media players generally only rely on a single CDN, because managing multiple CDN is too hard for them. In addition, by dividing volume between several providers, they may not get as low rates as they could by sending the entire volume through a single provider. Note that, even for video, some players are fine using transit and peering, especially those that don't need high QoS. This was certainly the case with Megavideo/Megaupload, depending essentially on Cogent for distributing its traffic until Megavideo's closure in early IDATE

83 Table 8: Various distribution choices for content providers Players Distribution role Details Amazon Own Cloud and CDN infrastructure, sold to third parties (AWS, CloudFront) Solution provider Still uses third party CDN for its video offer Facebook None Owns several datacenters Apple Uses several CDN (but mostly Akamai) Google (including YouTube) Microsoft Netflix Source: IDATE Own infrastructure Solution provider Own CDN and datacenter Own CDN since mid-2012 but no datacenter Uses Amazon's storage platform Still uses third party CDN (Level3, Akamai, Limelight) Google Like other Internet giants, Google has, from the beginning, massively invested in datacenters to hose its platform and content. With this storage capacity, Google then moved towards providing Cloud Computing services, requiring major storage resources. In parallel, Google acquired the video sharing platform, YouTube, in 2006, whose growing success required huge storage and distribution capacity, with redundant storage systems (even content stored in several locations for better security and optimised distribution). Google invested non-stop in datacenters based on its CAPEX plan. Google confirmed that most of its CAPEX is allocated to its IT infrastructure. Figure 88: Google's quarterly CAPEX spending since 2006 (Millions USD) Source: Google In 2008, especially faced with the success of YouTube, Google decided to build its own distribution infrastructure, "Google Global Cache" (GGC), which looks a lot like a CDN. This system optimises network infrastructure costs. GGC targets ISP and is implemented in the customer's datacenter, but is managed remotely by Google. The number of servers deployed depends on bandwidth demand from end users and their location. Nevertheless, telcos are very reluctant to implement Google technology in their networks. IDATE

84 Google's other interest is also in targeted advertising, providing detailed consumption analysis of subscribers to ISP. Targeted advertising is Google's core competency as well as its primary source of revenue (>95%) Netflix In the physical world, Netflix rents DVD for a subscription and mails them out. Customers select one or more movies from a catalogue on the Internet and receive the DVD by mail. They can watch them as many times as they want and return them by mail. Since January 2007, Netflix implemented a video content streaming service (movies and TV series) for viewing initially on a PC. Content streaming usage took over in 2010, so well that the company now sees itself as a streaming service that offers DVD rental by mail as an option. A study performed in January and February 2011 by NPD Group showed that Netflix represented 51% of the volume of movies downloaded or streamed in the United States, way ahead of Comcast (8%), DirecTV, Time Warner and Apple, each of whom had 4% of the total volume. Netflix had become one of the largest Internet services in the United States in terms of traffic in just a few years. Netflix has only worked with market CDN up to now. It was allied with Akamai before turning to Level3, also a transit provider. The share of traffic becoming significant resulted in a lawsuit between Level3 and Comcast at the end of But, in mid-june 2012, Netflix announced the release of its own CDN, called Open Connect. By becoming its own CDN, the company would reduce its Internet bill. Figure 89: Distribution of costs for the various CDN Netflix uses, before (on the left) and after (on the right) introducing Open Connect Source: DeepField But Netflix must also be able to expand with ISP. ISP can install the Open Connect appliances on their own networks or interconnect directly with Netflix. Netflix already has several exchange points around the world, not only in the United States. 5 In November 2010, Comcast, the cable operator and ISP, demanded that Netflix and its Tier 1 partner, the operator Level3 (acting as the CDN service provider for Netflix), pay for peering to route its movies to its customers. At the same time, Level3 accused the cable operator of imposing a new tax that penalized online video distribution services and placed operators broadcasting these videos in an unfavourable commercial situation. Comcast immediately denied any connection between this clash and Internet neutrality, and prefers to call this affair "a simple commercial issue". IDATE

85 Figure 90: Netflix's CDN peering points Source: Netflix Facebook In April 2012, Facebook had more that 900 million active users. Each day, users upload about 300 million pictures and send 3.2 billion "likes" and other comments. According to Sandvine, Facebook is the 11th most popular application on the Internet, which represents 1.5% of all traffic in the United States. On mobile sites, the social network is ranked third in popularity, which represents 10% of all traffic. Figure 91: One day's traffic on Facebook in North America Source: Sandvine When Facebook went public, it unveiled a few elements about investment in its architecture. It expected to spend more than 1 billion USD in network equipment by the end of There was no indication that it would build out its own CDN. IDATE

86 4.6. Positioning of other technical service providers involved Summary On the fringes, there are specialists like SkyTide, which provides reporting and analysis tools. SkyTide supports the federation of telco CDN and is heavily involved in providing a solution dedicated to this configuration. Other market players mainly include companies specialising in transparent caching, which, for the most part, work in partnership with the major equipment suppliers to provide a complete solution to operators, like Verivue who collaborates with Nokia Siemens with its CDN OneVantage platform. Today the transparent caching market is fragmented, with companies established in cache - PeerApp, Verivue, BlueCoat - as well as players coming more from network management and traffic optimisation in the mobile segment, like Qwilt, ByteMobile (acquired by Citrix in June 2012), Saguna and Ubiquisys, who integrate caching functionality in their solutions. In addition, the recent acquisition of Oversi by Allot in July 2012, a major player in traffic management, indicates strong interest by traffic optimisation players to enter the transparent caching market. Table 9: Primary transparent caching providers Type of players Pure Players Video optimisation Source: IDATE Players PeerApp Verivue BlueCoat Qwilt Allot (Oversi) Citrix (ByteMobile) Saguna Ubiquisys On a final note, video/media specialists (Anevia, Broadpeak, Edgeware), especially in the world of IPTV, are also investing in the telco CDN market with offers that are similar to telecom equipment suppliers SkyTide SkyTide has a specific position in the CDN market. The company provides reporting and analytics tools on Terabytes of data in real time for companies operating CDN. It allows an understanding of operations done across CDN. The data analysis process is performed continuously as it loads and makes it immediately accessible. Actually, the company differentiates itself from traditional data ETL (Extract, Transform and Load) as a data warehouse process. The data compression is up to 100 times greater than standard compression, requiring less hardware, reducing CAPEX. Figure 92: SkyTide role within CDN landscape Source: SkyTide IDATE

87 Telcos and ISP that launched their CDN can rely on reports and analytics to: Isolate and correct service quality issues Forecast peak traffic and provision capacity Justify pricing Ensure accurate billing. Advanced reporting and analytics can be supported by SkyTide. For instance, SkyTide provides traffic measurement by network node within a geographic area as illustrated below. Figure 93: Example of measures taken by SkyTide Source: SkyTide Furthermore, SkyTide has introduced analytics and reporting applications, designed to address the unique challenge of measuring network performance across a CDN federation. The solution gathers data across all content owners and federation members, while also providing secure, private views of data relevant only to each participant. SkyTide differentiates itself by providing a one stop shopping offer to content owners allowing them to get a unified view of streaming video performance through a centralised analytics & reporting exchange. The Federated CDN solution is able to manage different technology platforms and formats as well as display data through secure portals. IDATE

88 Figure 94: Federated CDN Source: SkyTide SkyTide partners with the main CDN providers such as Cisco, Juniper, Ericsson, HP, Broadpeak and JetStream. Leading telcos already rely on SkyTide solutions, including BT, Telefonica, Telstra, Telecom Italia and KDDI PeerApp Initially, PeerApp directly targeted content and CDN providers, but now also supports operators. According to PeerApp, CDN deployment and transparent caching by operators is needed. PeerApp has positioned itself as a software publisher specialising in transparent caching solutions on its Ultraband platform. The company doesn't want to be a CDN provider but a caching supplier to offload the network. Ultraband is a single platform that can both cache and accelerate content by progressive downloading, in HTTP as well as P2P. Ultraband accelerates media content broadcasting on the network while reducing infrastructure costs - up to 30% savings on a service provider's overall bandwidth The PeerApp platform includes caching software, storage and a management tool to deliver traffic statistics. According to PeerApp, its caching platform can manage traffic from various services with a capacity up to more than 20 Gbps, with up to 16 caches of 60 TB of storage each. To deliver a complete solution, PeerApp relies on major partners such as IBM and Dell for hardware, Sandvine, Allot and Arbor for DPI and Alcatel-Lucent as network equipment supplier. As shown in the figure below, Ultraband is integrated into the network using existing network elements, typically relying on a router or Deep Packet Inspection (DPI) equipment. Traffic is then directed by this element to the Ultraband platform, then relooped on this same network element, before being sent as if it hadn't been cached. As a result, according to PeerApp, there it requires only a small network configuration change by the operator, which is transparent to users because there is no major change in traffic flow. PeerApp doesn't store URL addresses but directly caches the content. IDATE

89 Figure 95: PeerApp's transparent caching solution. Source: PeerApp The output depends on the location of the caches, which can be located at peering points as well as lower in the network (almost to the last IP equipment), On fixed networks, caches are more often deployed at the service access equipment (BRAS), on the edge of the xdsl network (not at the DSLAM because of the feed), or at the CMTS cable modems for cable networks. For mobile networks, caches are implemented at the GGSN interconnection gateway. According to PeerApp, with a 24 Mbps ADSL connection, content can be provided in 150 seconds without caching, compared to 19 seconds with transparent caching, or content delivery 8 times faster. In addition, caching improves flow considerably. As shown in the graph below, during video streaming, flow for the same service without cache is 4 Mbps compared to 1 Mbps with caching. PeerApp's solution also improves QoE, particularly during peak hours or during traffic spikes (caused by sports events, for example). Figure 96: Cache improves flow Source: PeerApp According to PeerApp, more than 230 operators use caching solutions. Among these operators, Lime, the Caribbean subsidiary of Cable&Wireless use Ultraband, as well as the Israeli operator Pelephone. Ultraband is agnostic and can be combined with any CDN platform to manage both flow and OTT. IDATE

90 The company is working with most equipment suppliers like Cisco, Alcatel-Lucent and Ericsson. Nevertheless, PeerApp signed a key partnership agreement with Edgecast in September By combining Edgecast's CDN solution and PeerApp's Ultraband on a common platform, operators can handle managed and unmanaged content (even though the solutions are sold separately). Up to 70% of traffic can be cached as a result of this partnership. Figure 97: Amount of traffic that can be cached Source: PeerApp Qwilt Qwilt is company founded in 2010 by veterans from Cisco, Juniper Networks and Nokia Siemens Networks, initially positioned in video optimisation technologies. At first, Qwuilt provided distribution solutions for operators to control video flow on their networks. After two years, Qwilt developed a solution based on content classification and analysis to transparently detect video content transported in operator networks. Actually, its unified software solution, "Universal Video Delivery", combines transparent caching and the distribution of video analysis (classification, monitoring and optimisation). Several operators rely on Qwilt technology to distribute OTT content on the edge of their networks. According to Qwilt, with video files as close to the end users as possible, operators can decongest the backhaul and core networks (up to an 80% reduction in traffic). Qwilt is in discussions about its "Universal Video Delivery" solution with about fifty operators and content providers and is currently in test with five of them. IDATE

91 Figure 98: Qwilt's distribution solution Source: Qwilt Verivue Initially, Verivue developed and sold platforms to cable operators to distribute and manage large amounts of IP traffic, basically as managed services. To support OTT services, Verivue acquired CoBlitz in 2010, a company specialising in CDN directed towards telcos. In the context of an operator's CDN, the OneVantage platform targets all operators wanting to manage growing amounts of OTT traffic. TP, an operator in Poland, deployed OneVantage to build its own CDN to capitalize on the growing demand for online content by distributing premium and high quality content to its subscribers. Verivue's solution helped TP quickly and simply deploy transparent caching services. Finally, in 2011, Verivue integrated a transparent caching option in its platform as a software offer. As a result, OneVantage Transparent Internet Cache lets operators manage OTT traffic, while reducing peering costs, on the same CDN platform. Figure 99: Deployment of Verivue's tools Source: Verivue IDATE

92 BlueCoat Provider of Web security and WAN optimisation solutions, BlueCoat is now a major player in the transparent caching market. Alcatel-Lucent uses its transparent caching appliance as part of a partnership agreement. BlueCoat developed its caching and content distribution solution, CacheFlow, which speeds up content delivery (including video and large files) as well as controlling bandwidth. BlueCoat can reduce overall Web traffic by 40%. Its cache technology is based on proxy servers deployed in the network; the company estimates that content distribution time can be 2 to 10 times faster. In addition, CacheFlow uses a CachePulse appliance to verify updated cached content. As for deployments, CacheFlow solutions have been integrated into Cisco and Juniper routers, and in the load distribution switches of Brocade, Citrix, A10 Networks, F5 Networks and Radware. Figure 100: How BlueCoat's transparent caching works Source: BlueCoat Figure 101: How BlueCoat's CacheFlow transparent caching works Source: BlueCoat IDATE

93 5. Strategic Analysis 5.1. Regulatory issues and Net Neutrality Even the nature of telco CDN (ability to eventually manage from end to end) poses a key question in regulatory terms. Telco CDN is not a managed service in the traditional sense, in which ISP serve both as technical intermediaries and commercial intermediaries (to drive traffic and bill for it based on content use). Telco CDN are more technical intermediaries using specific resources (not at the level of the pipe but more at the level of the servers) but not a commercial intermediary. Telco CDN often comes up in debates on Net Neutrality, because in the end it can differentiate the quality of some services. Regulators don't want to limit development (unlike some in paid peering) for two reasons: Implementing a telco CDN does not necessarily imply de-prioritising the rest of the traffic on the open Internet. They use the same pipes and improvements in QoS come from caching (reducing unnecessary transport), which could actually improve the rest of the traffic. Prioritisation functionality could be implemented in combination with telco CDN (such as BT Wholesale's offer), but generally like solutions already in place (a classic problem for traffic management). Differentiation by telco CDN already exists on traditional CDN. Telco CDN pricing regulation is not necessarily needed because, unlike paid peering, there are alternatives, with (integrated) managed services and above all traditional CDN. Certainly, the quality is lower (the ISP manages his own network better to ensure the most efficient routing, and can rely on its caches), but traffic can continue to be delivered with a good QoS at normal and reasonable prices. In addition, traditional CDN prices, which are more or less competitive with telco CDN, are not regulated. Development of telco CDN by ISP can be done gradually, started first by the major ISP (it is clearly seen in the preceding section that initiatives are coming essentially from incumbent operators). These will have a stronger overall QoS, making it easier to manage heavy loads and make their Internet access more attractive. Other that the financial benefit gained by the ISP and savings in operating the network over a telco CDN, an ISP will have a competitive advantage over other ISP. This does not pose a major competitive problem. The other ISP can do the same thing because telco CDN deployment involves the core and collection networks, not the local loop. All operators, except for virtual operators, are in a position to do the same thing and optimise their networks using CDN (or other approaches). IDATE

94 5.2. Technical Issues To deploy CDN effectively, operators must have a complete understanding of traffic, content formats, as well as demographics. As a result, the primary issues telco CDN face involves the number of servers to deploy, their location and the associated capacity. Then they need to determine the content that needs to be cached, and the timeframe for holding cached content. To these main issues add using DNS redirection, as well as problems for operators to properly use analysis tools CDN sizing Actually, operator CDN can be deployed at several levels of their own networks, with the ability to go as deep in the network to be as close to the user as possible - typically all the way to the access network. However, to accelerate content distribution or optimise flow, caches can be deployed upstream at points of exchange as well as in the core network. Then they need to determine, based on the operator's strategy, the number of servers to deploy, because not everything can be cached. The business case must justify a CDN deployment and measure the associated CAPEX and OPEX. Existing CDN solutions from suppliers suggest consolidating servers. Using a ratio of 10:1, operators optimise the space, power and equipment cooling requirements by changing the number of servers to be deployed. Juniper Networks estimates that OPEX can be reduced by up to 90%. Figure 102: Optimisation by consolidating servers Source: Juniper What content can be cached? The operator CDN is responsible for assigning storage resources and quotas for each distribution node, as part of contracts with content providers. Based on distribution policies, and maximum cache time (Time to Live (TTL)) for objects determined by the operator, the CDN distributes content based on real-time content popularity. Overall, operator CDN are most efficient for regional and local content. JetStream's solution is to dynamically manage and distribute content based on the popularity of the objects and their location (network edge or core). IDATE

95 Figure 103: Storage based on popularity Source: JetStream As far as transparent caching, it can intelligently decide which content should be cached locally to optimise the network. The more popular the content, the lower it will be in the network (to be as close to the user as possible). The time to live for caching content is also one of the issues for telco CDN. If popularity is measured based on user request, operators are able to determine how much time the content remains in the cache, by setting TTL indicating content expiration DNS redirection issue CDN is based on DNS redirection, taking into consideration the user's location and determining the most appropriate content server closest to the user. Using DNS creates a double challenge for operators. request overload that can affect CDN performance given that each original request to the server is rerouted to the CDN. This routing can increase traffic to the CDN unnecessarily and involuntarily. Naming the servers. If a user has DNS service other than that of the operator, like GoogleDNS (independent DNS service), the CDN may not be able to determine the user's location, which could result in poor performance from the telco CDN. Also, if a telco CDN redirects the main server domain to another, the user who chose another DNS server may not be informed of the server redirection (unless the telco CDN has configured a content expiration using TTL on the main server). As a result, some CDN providers rely on alternative technologies other than DNS redirection to resolve this double problem. JetStream, for example, has developed a protocol that controls and guarantees redirection on the basis of the request. Each object and flow is constantly monitored geographically and for popularity Problems implementing analysis tools Generally, operators are not software savvy; this often results in problems implementing analysis tools, even though they are commercially available. In actual fact, consolidating logs to do reports is somewhat complex. IDATE

96 Multiscreen and mobile Network issues Video and the various services are now also used on mobile devices (tablets, smartphones). The CDN concept is still based on "fixed networks", with decentralised servers (edge servers) on the fixed network but not on the (existing) mobile network. So caching (one of the main components of CDN service) is nearly non-existent in mobile networks. Caching services offered by CDN providers on mobile networks use cache servers located on the fixed network, which are located on the edge of the access operator's core network. Nevertheless, LTE equipment suppliers are looking into integrating servers closer to the base stations, especially in high density areas, for very popular content, replicating the concept of saving fixed to mobile bandwidth costs (except for radio access). However, offers from providers of these mobile solutions are very focused on video technology and not on network architecture. The primary characteristics of mobile CDN providers are based on: technologies to resolve inherent problems in the mobile network, such as traffic shaping and adaptive encoding (described above); technologies for real time content adaptation of the image (resolution, the proper size and format for each type of device). CDN usage for mobile networks remains limited, with unimpressive results in terms of performance, because all congestion actually happens on the radio loop. Deployment of caching in mobile networks remains complicated (problems with location, can't be used on networks that are not totally IP) and very costly if deployment takes place at the base station. An easier solution combining traffic management and caching, integrated directly during next generation LTE deployments present the opportunity for deploying complete solutions. Figure 104: Ability to cache based on position in the mobile network Source: Everything Everywhere IDATE

97 QoS Content issues Beyond network issues, multiscreen distribution is another technical issue for managing content that needs to be adapted to specific devices (codecs, formats, operating systems, resolution, encoding, etc.). CDN must be able to offer solutions to manage these types of problems, based on cloud architectures combined within CDN. The QoS actually offered by telco CDN is still subject to debate On one hand, telco CDN need to prove their ability to offer QoS better than traditional CDN, who compensate for the network limitations (no end-to-end network management, in particular on the access network) with better software performance. The difference in QoS between solutions remains to be demonstrated, even if it should eventually materialize for large video files (even though traditional CDN are sometimes partners with telco CDN on technologies). On the other hand, the question is still open with IPTV integrated managed services. Will the quality between the two services be different, knowing that only the final run will eventually be different (for example, virtual channels)? Operators generally make more on premium paid content from the commission on managed services than through telco CDN; therefore, they are interested in offering different qualities Federation of CDN ISP CDN will only grow with the development of interconnection solutions between CDN. Only a few content providers are able to negotiate with all the operators in a country to ensure effective distribution, which is currently the process used for peering. Most players need a unique portal, which implies traffic distribution (with better QoS) to all of its users and not just to subscribers of a particular ISP. To do this, assuming that each operator has its own CDN, an ISP CDN should be able to redirect some of its traffic to the CDN of another ISP or to a traditional CDN CDN interconnection To offer the appropriate quality of service and simplify marketing 6, interconnecting traditional CDN providers, telecom operators and certain content providers could prove to be a good approach to resolve certain QoS problems. Figure 105: CDN interconnection concept Source: OCEAN project 6 Interconnect will make telco CDN offers (thus federated) more attractive for a content provider who wants to distribute its content to all subscribers in a given location. IDATE

98 Several interconnection concepts exist: telco CDN - telco CDN: interconnection could be made directly between telco CDN networks to offer "one stop shopping" solutions. Several initiatives are in development, some in standardisation organisations (ETSL) and in test by Cisco on the OCEAN project. traditional CDN - telco CDN: traditional CDN worldwide would provide global distribution and telco CDN would handle local delivery to the users. broker CDN platform: a broker CDN is an entity interconnected to one or more CDN who acts as a primary operator CDN for providing content. A special kind of interconnection would be offered by broker CDN. The first CDN provider (working directly with the content provider) would provide traffic to both consumers OnNet (it is a CDN itself) and to the CDN exchange platform, which is linked to other CDN providers. All members then interconnect to a centralized platform, to which all of the members interconnect. It could be that all of the preceding models could be assimilated into this one if a telco CDN or a traditional CDN plays the role of broker. Figure 106: Cisco's vision for CDN interconnection Source: Cisco Several players are looking towards the broker model or exchange platform model, but this solution is still pretty complex. It is hard to determine who is the best placed and the most likely to operate such a platform, especially for a third party without a CDN. Up to now, traditional CDN have been making the call. Therefore, the preferred model has been the traditional CDN - telco CDN, which will be described below. Note that Cisco has finished several interconnection tests between operators (see Cisco's profile). OCX A group of operators founded an entity called Operator Carrier Exchange (OCX) to interconnect their telco CDN networks to compete better against major traditional CDN, whose global coverage in terms of points of presence worldwide is decidedly larger than that of a single telco. Deutsche Telekom talks of its "Hub OCX" or mediation platform (see Deutsche Telekom's profile). OCEAN Project FP7 OCEAN ( in which IDATE takes part, studies various approaches to CDN federation and contributes actively to the standardisation efforts of ETSI. In addition to federation, the project offers innovative solutions on cache congestion control and encoding. IDATE

99 Source: OCEAN ETSI and IETF In addition, more global initiatives are pushed by ETSO for CDN interconnection (CDN-I). Figure 107: Federation of CDN initiations Source: TNO Broker concept A traditional broker would bring together a federation of CDN rather than a CDN interconnection. Actually, the CDN has no need to be directly linked (interconnected) with them. They interact through functioning by providing mediation by the broker. A CDN may not even be informed that it is part of a multi-cdn federated system. Brokers would be small, essentially targeting customers for whom video distribution it not a core competency (no expertise) or need one time event coverage. This is really a marginal approach. Customers are content providers or telcos and institutional sites. Table 10: List of the main CDN brokers Players Role Customer types Video (quality) Atanar CDN Aggregator Small players based in France +++ Streamakaci Vpod Streamer Audio and video event distribution Streamer Platform suppliers French media (radio and TV) Corporate Web sites (France) Institutional sites Content distributors Producers Corporate Telcos Advertisers (targeting TV) IDATE

100 Players Role Customer types Video (quality) VPS Hosting "Cloud" provider Corporate - (no reference to media) CDN Technologies CDN reseller Content distributors Corporate + MediaMelon CDN Aggregator Telcos and ISP Content providers +++ Conviva CDN Aggregator Content providers Corporate Advertisers +++ Source: IDATE Pure business of a broker is therefore based on buying of bandwidth capacity from one or more third party CDN. Then, the broker resells a service package for its customers, acting as intermediary between the content providers and the content distribution providers (telecom operators or CDN providers). The pure broker, also called a CDN aggregator, is a multi-cdn manager. Its strategy makes it easy for companies to choose various providers for content delivery. This choice is made mainly based on multiple parameters, like real time conditions in the provider's network, its geographical coverage, its expertise in specific applications. These selection methods are made possible by the measurement tools offered by the broker itself, using third party specialists like Cedexis Load balancing and multi-cdn management Cedexis is often called erroneously a CDN broker, but it doesn't aggregate or resell CDN products. Cedexis provides statistics on the performance of various CDN providers from the point of view of a lambda user on Web sites selected and tested (using a javascript tag) using several protocols such as HTTP,, RTMP and SSL. The indicators are then provided by AS (in real time for paid subscribers) as well as by ISP, by sharing all the data collected. Different types of traffic are tested, like video simulated by very large objects (greater than 1 Mb) and small objects. Figure 108: Illustration of the performance of various CDN on Deutsche Telekom's network Source: Cedexis IDATE

101 Using this tool (radar), a content provider can determine the best performing CDN and distribute its content between CDN. Load balancing is generally done between CDN taking into consideration various criteria like price, location, QoS requirements, performance in real time (like latency and time to download a web page). Functionality can even be integrated in the equipment supplier's or telco s routers. In addition, some CDN players also offer load balancing tools, like Level3 with its "Intelligent traffic manager" solution. Using this product,, customers can decide in real time the best way to deliver their traffic between their CDN existing providers based on their existing contracts. But Cedexis doesn't manipulate or redirect traffic. The content provider / service must manage that itself. Its customer list includes several content providers that have adopted a multi-cdn strategy, like France Televisions in France for example Developments for federation Equipment suppliers and operators are relatively optimistic about the early establishment of CDN interconnection. But it all depends on interconnection between ISP. Though promising, it seems like it s not enough to target the global market. Actually, some ISP do not have CDN. The issue rests mainly, in the short and medium term, on the CDN interconnection between telco CDN and traditional CDN, which will continue to capture the largest amount of traffic from Internet players. Without interconnection, this model won't gain the size needed to make the operator's investments pay off. CDN players seem to be gradually moving towards this solution, even though they are only partially and indirectly embracing it, seeing the federation as a means to standardization or monopoly: they are certainly ready to ensure interconnection happens by pushing their technology into the ISP network (and selling it). Remember that, to resolve the CDN - telco CDN interconnection problem, the ISP can try to push transparent caching, getting around commercial contracts, while traditional CDN already have these contracts Economic issues Network investment The actual profitability of telco CDN remains uncertain because it depends on several correlated factors: investments (in deployment), cost savings, revenue (share of traffic captured, cache efficiency 7, unit price of telco CDN). The amount of traffic actually captured is essential. In the event of a significant disagreement with Internet players or with CDN, they won't get any traffic and the investment would be worthless. Telco CDN also provide managed services as well, given that deployments are generally hybrid. This risk is significant, with the interest in transparent caching, which can capture a significant amount of traffic even though performance may be less than that of a telco CDN. Managed services hold a particular position in a telco CDN. The can partially finance a CDN, but the telco CDN can offer relatively competitive OTT distribution compared to managed services, because they increasingly rely on the same technical infrastructure. Telco CDN can therefore eat into managed services revenue, which is key to ISP revenue and network financing. 7 Caching performance, the percentage of files correctly identified for caching IDATE

102 Economic transfers between players CDN interconnection poses unresolved economic problems on sharing the value generated, in particular between traditional CDN and operator CDN (the telco CDN - telco CDN normally causes less of a problem because it can be assimilated into peering between ISP). Several business models have already been put forth by several players (telecom operators as well as telecom equipment suppliers) with cascading mobile as the base model. CDN interconnection with traditional CDN will be required at least in the medium term because they carry most of the traffic. But traditional CDN are not necessarily interested in interconnecting because that could result in a drop in their revenue and their margins. Operators must find better incentives for the CDN to cooperate. The "status quo on free peering" model (by establishing telco CDN) In this model, the content provider pays the traditional CDN, who keeps all the distribution revenue even though it peers for free with the telco CDN. The operator does not receive any revenue from the upstream chain (content provider and traditional CDN) and only receives revenue from the end user who pays a basic subscription (generally unchanged because of competitive pressure from other ISP). It s an extension of free peering, whose formula is in question, but remains a potential option based on savings from caching. One variant is sometimes brought up in which the content provider pays the traditional CDN only for content actually distributed by the traditional CDN (i.e. If the content is cached by the traditional CDN). In this case, the operator also doesn't receive any revenue (except from the retail subscription), but the bill for the content provider would be lower. The cascade model This configuration is based on the base model ("status quo"), but this time there is a business transaction between telco CDN and traditional CDN that looks like paid peering. Actually, the content provider pays the traditional CDN for all the distribution. In return, the CDN pays the telco a distribution royalty for the content sent to its subscribers. This creates a cascading system with payments along the entire value chain. IDATE

103 American example of Dr Peering This model assumes that revenue goes to where the infrastructure is most often used. Tata receives the largest share of revenue for its main distribution contract, and AT&T and Cablevision receive additional revenue for delivering the content. The interests and incentives here are identical. Tata, AT&T and Cablevision all receive additional revenue, but only if content is successfully delivered. But if problems come up along the way, all are incentivised to correct them in order to generate additional revenue. Figure 109: Cascading model revenue flow Source: Dr Peering In a live streaming scenario, all the participants take part in the costs of delivery, but in a downloading scenario (static movies), content may be cached. However, when the content is cached in Cablevision s network for example, Tata s and AT&T s load level is very low, or potentially zero. Model based on cache server usage or "Pay per cache" The closest cache server holding the desired content is the only one to send content. The request is sent first to the telco CDN cache that delivers the content if it is in its server. If not, the request is sent to the traditional CDN which sends the content. In this scenario, the content provider pays the traditional CDN and the telco CDN for the traffic actually delivered (the traffic actually coming out if its servers) by each respectively. Revenue is therefore proportional to the number of requests on the cache servers of the respective CDN (telco CDN and traditional CDN). In this way, all the CDN players (traditional and telco CDN) are paid for the first distribution (of the same content), then only (and respectively) when their cache servers are used or "requested". This is a business model based on actual cache server usage (and the actual network used), and so it's called "pay per cache". IDATE

104 Figure 110: Illustration of the "pay per cache" model Source: Dr Peering The extra QoS model In this configuration, QoS is paid by the content provider and not by the end user (who still pays a basic subscription and for (paid) content that it views). However, the content provider pays the traditional CDN for all of the distribution (not only if the content is actually delivered) and also pays the telco CDN for QoS. The Web service/content provider pays both types of CDN for complementary services, limiting the payment to telco CDN to only that content that truly needs higher QoS. The content transits through both CDN. This model is generally considered to be marginal and seems complex to implement and measure. We won't discuss it further Decisions on setting price Operator CDN must generate new revenue for content providers, directly or through an intermediary CDN if interconnected. However, the operator must make several decisions, both on deployment and on price. Server placement and impact on costs The first decision is technical: deploying servers in the access network. This decision is obviously based on the overall level of traffic and the unit cost of the servers. The servers save on transport but they cost a lot if they are close to the subscriber (more servers smaller in size). The more traffic an ISP can attract to its CDN, the more business sense it makes to move its deployment closer to the local loop. But, it's uncertain that going too low in the network (at least in a fixed network) would be a good idea, given overcapacity in the local loop. On fixed, where most problems arise, operators, especially the telco CDN, are looking to reduce collection costs. The game may be different on mobile, where the issues are focused on the local loop. IDATE

105 Savings versus new revenue The true business decision is the price for telco CDN (in this case, the cascade and pay per cache models), subject to the return on investment, which is directly related to the volume of traffic captured. Telco CDN could represent significant network savings, inducing ISP to capture as much traffic as possible, for a very low price (or basically no price). In many of the current ISP simulations, telco CDN first of all represents a cost savings. However, savings could be compensated for (in absolute value) by the increase in traffic, which results in an efficient and less expensive telco CDN (compared to a traditional CDN). Telco CDN can also be seen as a source of financing for the network and a signal to content providers, who already often pay a CDN for service partially based on traffic volume. The overall price could then be aligned with traditional CDN, or a little more expensive (due to the enhanced QoS). This risk is to only attract a little traffic, which limits network savings, because the competition from traditional CDN would remain strong. Price levels The price of a telco CDN could then be based on two reference prices (and of course its cost): traditional CDN prices: that can be seen as a competitive solution, with potentially a differential for QoS; paid peering prices (when this solution is available): for less demanding solutions, paid peering could suffer. ISP have several options (which could result in possible complaints from the players), knowing that it generates savings on traffic transported: no price (status quo model), to capture more traffic. low price, to induce all players to migrate from paid and free peering, especially traditional CDN. This is based on a cascading or pay per cache model. medium price, to migrate some CDN customers directly to telco CDN. The higher the price, the less traffic will be captured, because the traditional CDN will try to stay with paid peering. These also are based on a cascading model, especially pay per cache. high price, targeting players looking for very high quality of service. This model is based mainly on pay per cache and extra QoS. ISP will choose various strategies based on the competitive landscape, as well as on their ability to reduce costs and payment models, and an interconnection model (allowing it to also capture more or less traffic). IDATE

106 5.5. Relationships between players Recall that telco CDN could benefit the primary players in the chain, besides the traditional CDN: content providers, who see usage grow (due to better QoS) and therefore their revenue; telcos, who can reduce their costs by caching traffic and combining their OTT distribution with managed services resources. consumers, who get better quality of service. Changes in relationships between traditional CDN and operators With telco CDN, the pressure is on traditional CDN, who are already under pressure from transit providers evolving towards CDN. The battle is already ferocious between CDN, resulting in consolidation and acquisitions by equipment suppliers. Though CDN initially viewed telco CDN as a direct competitor, they are now trying to see them as an opportunity to evolve their offers towards licences and managed services, in competition with telecom equipment suppliers. They also have an interest in developing this market to no longer sell services to content providers but technological solutions to operators. The development of telco CDN is very pertinent and not easily replicated by alternative solutions beyond the access network for the most demanding services (like connected TV) and mobile services, even though mobile CDN offers for video are still inadequate. Traditional CDN, unable to penetrate fixed or mobile access networks without relying on operators and equipment suppliers, also have an interest in expanding telco CDN for some services or in certain circumstances. A true federation makes them more and more marginal. They must continue to drive activities in this direction (especially for the largest ones like Akamai), continuing to support federation based on their own solutions. A neutral position for media players Media players, in particular large ones, particularly like telco CDN, because they already have contracts with several CDN. It's the opposite for smaller media players, which are working with a single CDN. Telco CDN is less relevant, only covering a portion of the population (unless a part of a large federation). Telco CDN is an additional option for them, in addition to other possible services. Complex relationship with Internet players Then there is the ticklish issue with the major Internet players. Like the leading CDN, they are looking to push their caches into the operators' networks. However, they have little interest in using the operators' CDN (or at least paying for it, like for peering) and some of them seem content with lower quality. Unfortunately, they are generally the largest source of traffic. Without a CDN contract, it is impossible to cache their traffic (at least is some areas in Europe). Operators could lose out on a large amount of traffic, as has already happened with traditional CDN, except by relying on transparent caching. The double objective of lowering costs and generating revenue is difficult to achieve without involving these few Internet players. Those with the highest premium content, like Netflix, however, are in danger of upsetting everything, even though major political-strategic blocks are being laid based on double positioning of the operators. Telco are looking to actually become CDN providers for all content providers, while offering content services to users (through managed services for OTT). Operators are now potentially both wholesale providers and retail competitors to Internet and media players. Internet and media players are often the operators' partners. Telco CDN is therefore an alternative to managed services for these players. However, Internet players are not partners with operators for managed services and are therefore pure competitors. Given the conflict between Netflix and Comcast about the Comcast-Level3 case, there is little evidence that Netflix can collaborate with Comcast or AT&T, who want to push their own video services. The game could be different for other services for which operators have no retail offer (e-commerce for example) or the only alternative (mobile). IDATE

107 The Comcast - Level3 - Netflix case In November 2010, the cable operator and Internet Access Provider, Comcast, demanded that Netflix and its Tier 1 partner, the operator Level3 (providing CDN services to Netflix), pay for carrying movies to its customers. Then, Level3 accused the cable operator of imposing a new tax which penalized broadcasting online video services and placed operators that broadcast these videos in an unfavourable competitive position. Comcast quickly denied any link between this clash and Net Neutrality, but prefers to call this affair a simple commercial difference." In fact, Netflix and Level3 feel that Comcast wants to promote its own video platform, Xfinity, by ensuring better quality of service. Actually, the cable operator lost more than 620,000 customers for its video service in the first nine months of 2010, notably due to similar offers such as Netflix (7.99 USD per month for unlimited viewing). At the end of April 2011, Netflix even passed Comcast in terms of subscribers (23.6 million compared to 22.8 million). As far as Netflix and Level3 are concerned, Comcast violated the principle of "Internet neutrality", which states that flow on the network cannot discriminate. They called upon the American telecom regulator, the Federal Communications Commission (FCC). At the end of February, AT&T and NCTA asked the regulator for a clarification on its position and definition of Net Neutrality. The telecom operator and the American association of cable operators feel that the FCC's language on the open Internet only covers the "last mile" (the part of the network closest to the subscriber) and its non-discriminatory access, and not the backbone core (on which the litigation between Comcast and Level3 rests) and its transit/peering commercial concept. Nevertheless, Level3 argued that the conflict wasn't about the backbone but the access network, because Comcast wanted Level3 to pay for carrying flow to its subscribers. Level3 added that there was no peering contract between Level3 and Comcast. The debate is therefore about the definition of the boundaries of the FCC rules, either the last-mile or the network core. IDATE

108 Retail vs. wholesale Finally, note that this double positioning of operators puts in competition their retail business (example: VOD) and their wholesale business (telco CDN) whose levels of profitability are logically very different. Developing telco CDN could be to the detriment of its managed services. However, margins on retail are generally much higher. Operators don't actually get revenue directly from the UGC free videos or catch-up TV they transport. Telco CDN could generate revenue from its services for OTT players, while these services are only costs at this time. Transparent caching remains an attractive alternative. Table11: Usage assumptions Direct net revenues with telco CDN Direct net revenues with Managed Services Indirect net revenues with Managed Services Source: IDATE 5.6. Market Potential Retail and wholesale revenue from video distribution UGC Catch-up TV premium VOD SVoD Average YouTube Average monthly 5 EUR TTC per Netflix-like monthly usage BBC consumption movie subscription (2.5 hours per (4 to 5 hours per (10 EUR TTC) month per user) month per user) to to 0.20 EUR EUR per user per user (up to 0.10 EUR (up to 0.35 for with connected connected TV) TV) 0 EUR Negative (telcos paying for rights) Around to 0.10 EUR per sub. (not per user) 0.03 to.015 EUR per video Yes (marginal) Yes (up to 1 EUR?) None (except for bundled videos) 0.2 to 1.2 EUR per SVOD sub EUR 2 to 2.5 EUR per SVOD sub. None (except for bundled videos) The technical and commercial success of telco CDN solutions is uncertain. Technically, there are several challenges but they will eventually be resolved. Operators need to compensate for being technologically behind traditional CDN to prove that QoS is better with their own solution. This seems to be possible in the short term, by focusing on video and using third party specialists. Commercially, operators face a dilemma. Their CDN, even shared with managed services, is costly (in particular granular deployment and/or usage of high end servers). They can offer very low prices to be competitive and attract the most traffic, generating more savings. In contrast, they can generate more direct revenue with higher prices, but may not capture as much traffic (and therefore fewer savings). ISP CDN will only grow with the development of interconnection solutions between CDN. Only a few content providers are able to negotiate with all the operators in a country to ensure effective distribution, which is currently the process used for peering. Most players need a unique portal, which implies traffic distribution (with better QoS) to all of its users and not just to subscribers of a particular ISP. To do this, assuming that each operator has its own CDN, an ISP CDN should be able to redirect some of its traffic to the CDN of another ISP or to a traditional CDN. Each CDN uses very different equipment, which implies that standardisation is needed to make the CDN interoperable over several elements (TTL, caching ratio, etc.), which will take another two to three years, despite initiatives already underway, notably IETF (CDN-I Working Group) and Cisco (tests with several operators). IDATE

109 Small market share in the short term In the short term, operators will only be able to handle a small part of the CDN market for video with few value added services for the major media players (but no small media players or major Internet players). Based on low expectations for growth (for operators), decisions will need to be made on solutions offering significant cost savings and hybrid synergies. This could induce operators to put forward relatively cost competitive rates. Growth in cost savings will certainly be stronger that in new revenue, even considering adding small media players later and assuming a possible growth in traffic due to better QoS from telco CDN and the ability for media players to directly target connected televisions. Long term potential In the long term, it is probable that operators will evolve towards more complete CDN offers targeting the more profitable types of other segments (such as e-commerce) with value added services and unlimited mobile offers, approaching true mobile CDN. Operators should also accelerate market expansion in areas poorly served by CDN, benefiting from their local presence (network, sales teams, etc.). A market more important strategically than economically The telco CDN market should grow from 4.7 million EUR in 2012 to 360 million EUR in 2017, or from 0.2% to 5.5% of the total market. Despite this great growth rate (+138% per year), telco CDN will remain relatively minor out five years and even out to 2020 (a linear progression based on a boost from mobile resulting in operators capturing only 15% of the total market). But it will bring true growth to mobile. It remains strategic for operators in terms of QoS and cost savings (revenue is only a part of the complex equation for operators). In any event, revenue from telco CDN is minuscule compared to traditional retail revenue from selling broadband (on the order of 0.2% of the total telecom market by 2020). Telco CDN is only a (significant) enabler for establishing stronger strategies based on both sides of the Internet markets. Figure 111: Telco CDN market, Source: IDATE IDATE

110 Research Catalogue Reports & Services Networks & Telecom Networks & Equipment Telecom Strategies Mobile Broadband / FTTH Satellite TV & Internet Internet Services TV & Video Digital Content Digital Home Video Games

111 2012 Research Catalogue IDATE presents its Market Reports Programme for 2012 which constitutes a natural extension of the work performed by our teams of analysts, as well as our ongoing investments in information and monitoring systems for player strategies and markets. More than just a catalogue of publications, this constitutes a concrete manifestation of our drive to create a unique tool geared towards understanding and monitoring the Telecom, Internet and Media sectors. 3 pivotal collections for an easy navigation between the different reports and services: Watch Services Continuous watch on high-potential markets including trackers, insights & live sessions with lead analysts Market & Data reports Exhaustive market reports: trends, players, actuals & forecasts of markets by segment and by country Innovation reports Qualitative approach of prospective issues

112 Contact us: Isabel Jimenez - i.jimenez@idate.org - Tel: > More details online - Contents, Methodologies, Summaries on