CTO s cloud(y) agenda Clouds on CTO s heaven? Cloud computing is one of the most advertised technological trends and its discussion mainly focuses on cost efficiency aspects of IT infrastructures. This is usually the CIO s domain, but what is actually in it for the telecommunications CTO? Among the pain points on the CTO agenda are the flat rates models that cannot cope with the increasing network traffic demand. Cloud computing actually makes the situation even worse because of the additional network traffic that is generated for transporting data to and from the cloud. For the pure network operators the associated cost pressure is a challenge that adds to the already tough competition coming from the device manufacturers and the web players. As one solution could be to jump on the cloud computing bandwagon, this paper critically investigates which parts of the telco network are suitable for cloud implementations and which are not. At the end of the day the main question is whether the potential cost savings realized by cloud computing can compensate for the associated expenditures in the access and core networks. We make ICT strategies work
Setting the scene Cloud computing is today one of the most advertised trends in ICT service provider communities. What has not yet been sufficiently explored is the extent to which cloud computing can be profitably adopted for a telco s network near functions. Two key aspects that have to be clarified are 1) the areas in which you make use of cloud computing concepts and 2) what are the telco-specific technical benefits and challenges of cloud computing? These questions will be addressed by analyzing the technical aspects of cloud computing with respect to the typical telco network architecture stack. Other directions for analysis, such as the potential usage of cloud computing for the telco s IT systems (OSS and BSS) or the opportunities open to a telco to become a cloud service provider itself, are valuable approaches but not in the focus of this article. What are the key technical concepts of Cloud Computing? As a basis for further analysis the main cloud computing technical concepts will be explained here. First a clear message: Cloud computing is not a technology; it is a business model for the flexible provisioning of IT services. Traditional sourcing concepts rely on the use of dedicated hardware and licences. However cloud computing services are characterized by pay per usage, on-demand self-service, sharing of IT resources, and rapid elasticity. 1 Cloud computing is typically offered in three flavours: Infrastructure as a Service (IaaS): Provides infrastructure components like computing power (CPU), storage, firewalls, load-balancers, etc. in a virtualized environment. Example: Amazon s Elastic Compute Cloud (EC2) Platform as a Service (PaaS): Additionally provides operating systems, database and middleware components in an integrated runtime environment, e.g. for developing, testing and deploying applications. Example: Google App Engine Software as a Service (SaaS): Provides specific business functionality, typically based on standard application software. Example: Salesforce CRM Furthermore a fourth alternative can be observed when taking a closer look at telco operators. as a Service (NaaS). Bandwidth on demand is a data communication technique for providing additional capacity on a link as necessary to accommodate bursts in data traffic 1 See also: DMR 2/2010 - The Heavens Open - How cloud computing is growing up, Martin Jeske, Thorsten Claus; The Future of Cloud A Roadmap of Technology, Product and Service Innovations for Telecoms, Thorsten Claus, Daniel Kellmereit, Yasmin Narielvala (2010) Detecon International GmbH 12/2010 2 www.detecon.com
Cloud Computing benefits & drawbacks Today many CIOs consider cloud computing as a way out of permanent cost pressure on IT infrastructure and services. Benefiting from the cloud service provider s economies of scale is thereby the key factor. In some aspects cloud computing can indeed help. Cost structures shift from investment costs (CAPEX) to operating costs (OPEX). Budget allocation to accounts gains more flexibility due to usage-dependent billing, but also increases uncertainty in planning. New requirements or organizational changes that imply a new scaling of resources can be easily adopted using a dynamic allocation of resources and thus reduce time-to-market. But there are also drawbacks. One major aspect is the proper treatment of the security of shared resources. Carefully evaluated security concepts have to be considered. A second aspect is the possible lock-in to the cloud service provider due to currently missing standards, interfaces and protocols. Therefore, one of the prerequisites of any cloud implementation should be a clear and agreed procedure for data and code migration. Finally, the centralization of IT infrastructure implies more network traffic between users and cloud providers. Certain cloud computing drawbacks become even worse for telcos. For example, the shift of services into the cloud has a foreseeable impact on the access and aggregation networks in that additional capacity is needed for transporting the information to and from the cloud. In the context of corporate customers this can be partially compensated for within new contracts for site connections. In the consumer market we see another trend. Mobile apps are invading the markets, but the related revenues are largely collected by the apps/service providers and not by the telcos. On the contrary, these apps just generate significant extra traffic, which is making most operators reconsider their mobile flat-rate models. The operators have two ways out of this dilemma: either they must invest massively in the expansion of the network infrastructure, or they must face up to quality of service problems. The main question is whether the savings introduced by cloud computing will compensate for the additional costs of the network expansion? How are Cloud Computing concepts suitable for telecom operators? When focusing on the typical network structure of a carrier grade telecommunications service provider there are essentially four layers which could be relevant for cloud computing implementation (Figure 1): 2G/ 3G/ 4G WiMAX Access Backhaul Aggregation Core DSL/FTTX User Devices and Clients Access s Core IP s Core s Service Execution and Production Applicability for cloud computing Fig.1: layers potentially relevant for Cloud implementation Detecon International GmbH 12/2010 3 www.detecon.com
User Devices and Clients Regarding user devices and clients telcos may consider making use of the SaaS or PaaS concept for reusing software functionalities from the cloud. This would relax the devices in terms of required computing power by using thin clients that would retrieve and display the service information being processed by the cloud resident software. However, a thin client implies a significant increase in the traffic to be transported by the access and backbone IP networks. This adds to the existing problem of traffic explosion resulting from media consumption and smart phone usage, a subject which has already been addressed in various studies 2. Access s The foreseen bottlenecks are the access networks. This is due to the physical limitations of the access medium, which cannot be extended infinitely. Even though modern modulation techniques can improve this matter, the physical limits dictated by the transfer medium s characteristics cannot be changed. The access network is a network layer that is not recommended for cloud implementation. One critical argument is the natural proximity of access networks to end-users: both fixed and mobile access networks terminate close to the end-users and cannot be centralized. An innovative way out of this situation could be the development of new cooperation models between telcos. One example could be access network resource pooling between operators, or frequency spectrum sharing, which would leverage the cost benefits of dynamic resource sharing. Core IP s The SaaS concept for user devices would also impact the IP core networks by means of significant additional traffic. However, IP networks are by nature highly scalable and have a central positioning in the operator networks. IP routers are difficult to conceive as centralized, cloud based implementations, but a similar concept, namely equipment sharing, could deliver equivalent results. To reduce costs, network operators could partner to share expensive routing equipment and to optimize the traffic within the core networks. Service Execution and Production At the service execution and production layer it is possible to apply all three versions of the cloud computing concept, due to the characteristic feature of using mainly scalable, shared platforms. For example, IaaS and SaaS could be deployed for service execution and operation, and PaaS for service development and testing. Another argument for the wide applicability of cloud computing concepts to the service execution and production layer is that it is by nature centralized within the telco architecture, and not distributed, i.e. to the end user s proximity. Therefore, it can be concentrated in a data center and dynamically shared, making it a strong candidate for cloud computing. 2 DMR 01/2009 - Speed Is (No) Magic, Dr. Hans-Peter Petry, Daniel Schultz, DMR 02/2010 - Between Scylla and Charybdis, Dr. Hans-Peter Petry, Dr. Stefan Schnitter, George Salisbury Detecon International GmbH 12/2010 4 www.detecon.com
Implementing the service execution and production layer in the cloud implies certain cost savings, but the operators should not hope that these will completely compensate for the expenditures in the access and core networks. Conclusion & Recommendation Cloud computing is a trend that most telcos have embraced due to the increasing cost pressure on the telecommunications infrastructure and the benefits provided by cloud computing s dynamic, highly scalable network architectures. In general, any network resources or equipment that can be shared dynamically and that can be centralized in a data center are theoretically candidates for cloud computing. Nevertheless, due to the absence of standards it is recommended that the migration of any potential data and code from the cloud implementation back to the operator or to another cloud provider must be clearly defined and contractually binding. The network components that are located in the users proximity, such as access networks or traffic aggregation equipment, are not suitable for cloud computing. On the other hand, central equipment that is not directly connected to the user environment is a strong candidate for a cloud implementation. In a nutshell, the extent to which the cost savings introduced by cloud computing can profitably compensate for the extra expenditure made necessary in the access and core networks, due to the significant traffic increases, must be analysed on a case by case basis. Moreover, telcos need to weigh up the above-mentioned expenditures against the effort and costs of migrating existing operational services to the cloud. Telco CTOs should also innovate regarding cooperation models between operators, as these could provide a solution to compensate for the costs arising due to the constant increases in access network traffic. The feasibility of concepts such as access network resource pooling or frequency spectrum sharing should be further investigated, while taking today s regulatory situation into account. The authors Valentin Titica Valentin.Titica@detecon.com Christian Möws Christian.Moews@detecon.com Detecon International GmbH 12/2010 5 www.detecon.com