Grid/Cloud Computing Interoperability, Standardization and the Next Generation Network (NGN)

Transcription

1 Grid/Cloud Computing Interoperability, Standardization and the Next Generation Network (NGN) G. Caryer Telecommunications Management Consultants Ltd, UK J. Gallop Visitor, STFC Rutherford Appleton Laboratory, UK J. Grabowski University of Göttingen, Germany T. Rings University of Göttingen, Germany S. Schulz European Telecommunications Standards Institute, France I. Stokes-Rees Spmetric Ltd, USA T. Kovacikova University of Zilina, Slovakia Abstract For telecom operators, the future lies in converging fixed, mobile and data services onto the Next Generation Network (NGN). This paper discusses the relationship between grid and cloud computing, identifies gaps and overlaps in existing standards and identifies how grid and cloud technology could be exploited to improve the efficiency of NGN resources and to offer new data services to consumers. This will enable telecom operators to manage their resources in a dynamic and optimal way by a single platform. This paper describes the approach taken by the European Telecommunications Standards Institute (ETSI) Technical Committee for grid computing (TC GRID) to identify gaps and overlaps in grid/cloud computing standards and to support the integration of grid/cloud computing with the NGN architecture. Keywords: grid; cloud; NGN; Next Gerneration Network; interoperability I. WHAT ARE "GRIDS" AND "CLOUDS" The vision of grid computing is to create a universal source of computing power. The term grid is chosen by analogy with the electric power grid. Grid or grid computing is defined as "a system that is concerned with the integration, virtualization, and management of services and resources in a distributed, heterogeneous environment that supports collections of users and resources (virtual organizations) across traditional administrative and organizational domains (real organizations)" [1]. Grid computing has primarily been adopted by researchers in the public sector to meet the computing and data demands of their applications. Grid computing provides large scale federated computing systems for this community. For their unified access, common interfaces need to be developed. Grid computing aims to provide a collection of service interfaces which may be implemented and deployed by a range of providers. A user of, for example, the Enabling Grids for E- science (EGEE) [2] grid may, without necessarily knowing it, use resources in France on one day and a differently implemented resource cluster in Bulgaria on another. In this case, for example, the computation might have been split between independent resource providers by multi-parameter sweep. The EGEE grid and other large grids are able to offer a large collection of services on heterogeneous resources to diverse organizations. The diversity within a grid is a benefit but also adds complexity to the process of establishing relationships between service level agreements (SLA) and charging policies. Grid computing offers the concept of virtual organizations and supports single multi-enterprise applications involving multiple service providers. For example, this is demonstrated in a number of projects in the 6 th European Union Framework [3]. Users and resources are part of a federated network. A grid provides a security framework for identifying interorganizational entities (both human and electronic) and managing access to data and inter-entity services. Grid technology continues to be dominant in the scientific computing environments of the public sector. The reasons are the collaborative nature of scientific work and the need to manage existing data sets and computing resources across organizational boundaries. A more advanced state of interface standardization within grid technology will allow some degree of choice between various software and hardware systems. In contrast, cloud computing offers a sense of flexible, abstracted resources featuring scalability, a pay as you use model, reliability, performance and client-initiated variation of resources accessed.

2 Cloud computing originated in the private sector. It is based on virtualization technology and provides efficient operation for massive data centres. Therefore, it became an alternative to costly and operational challenging in-house management and operation of data centres of many businesses. The term cloud computing originated from the Amazon Elastic Compute Cloud (EC2) that is the first large scale, commercial commodity computing service and started in August Similar services have followed from other Internet giants such as Google, Yahoo and Sun. Amazon itself has followed up with storage services the Simple Storage Service (S3) and SimpleDB [4]. A cloud instance is typically offered to a single user at a time or possibly to groupings of clients that are normally part of the same organization. Any further sharing of the resource must be managed by the end user(s) via interfaces presented by the cloud provider. Cloud computing offers a solution to the problem of organizations that need resources (computing, storage, or network bandwidth) either quickly or with a highly dynamic level of demand. But when operating in steady state at or near full capacity, cloud computing might be more expensive than direct ownership of computing resources [5]. Cloud computing provides a simple model for access and development. Users can lease an arbitrary number of processors and storage for an arbitrary length of time and the charging is directly related to their usage. User-requests are satisfied by allocating a partition of the underlying physical infrastructure and instantiating virtual interfaces to the allocated physical resources that are only controllable by the requesting user. Currently, there is limited support for coordinated resource access. However, if users are able to allocate virtual resources on demand, they can dynamically scale their partition by requesting resources. Deploying data and applications into a cloud environment limit the organization to a single cloud provider or require duplicated effort because the deployment process needs to be repeated for different cloud environments. The current state of the art favours cloud computing for a single organization that can deploy commercial applications into a cloud environment. The dynamic provisioning of storage, computing power, and network bandwidth allows rapid scaling for intensive utilization either directly by the organization or by the public via Internet-based interfaces. During 2008 and 2009 commercial interest shifted from grids to clouds, with the availability of several on-demand compute and storage resources. The similarities are that both aim to provide access to a large compute (CPU) or storage (disk) resources. Beyond that, a cloud utilizes virtualization to provide a uniform interface to a dynamically scalable underlying resource, with the intention that the virtualization layer conceals physical heterogeneity, geographical distribution, and faults. The cloud environment only provides direct support for single user or single organization access, and current models typically have a high cost to integrate computing, data, or network transfers from outside of the cloud. In this document, the term cloud or cloud computing is considered to describe a rapidly provisioned infrastructure (e.g., compute, storage, network resources) that supports dynamic scaling with uniform interfaces to these resources. Both, clouds and grids, offer a vision of facilitating access to large data and computing resources. Many of the ideas which now are covered by the term cloud are part of the grid vision. Although, the definition of cloud computing varies [6][7], the following key concepts have been identified: dynamic, scalable, rapidly provisioned and virtualized. The definition refers to these properties in the context of foundational infrastructure resources. In contrast, grid computing assumes that a physical infrastructure is present. It focuses on a middleware service-oriented layer that emphasizes the key concepts of federated, distributed, and heterogeneous resources. None of these are visibly in cloud computing environments. Their distributed and heterogeneous properties are addressed by a cloud provider through virtualization of their physical infrastructure. However, a cloud can be realized on top of a grid. Grids and clouds provide solutions for different areas. In a cloud, potential users demand economical access to reliable, scalable computing and storage, whereas the difficulties of their realization and management are solved internally by the cloud provider with their expertise and economies of scale in their massive shared data centres. Grid provide solutions to situations where potential users are faced with a data deluge, distributed user communities, and federated computing centres that all need to interoperate in a secure way, and provide mechanisms for extensibility. Grid technology aims to provide software and services to achieve those ends. Currently, cloud systems only provide the foundation onto which applications must still be deployed and managed. Grid computing addresses different issues around federated interoperation of computing facilities, security, shared data management, application deployment, system monitoring, and application or job execution. At the present, these issues are normally not considered within cloud computing offerings. II. STANDARDIZATION AND INTEROPERABILITY In order to provide a basis for its work of addressing issues associated with the convergence between IT (Information Technology) and Telecommunications, the European Telecommunications Standards Institute (ETSI) and its Technical Committee for grid computing (TC GRID) produced a number of Technical Reports (TRs). The ETSI TR captures the current state of grid and cloud technologies and identifies the key stakeholders including standards making bodies, research projects, production grids and other initiatives [8]. Additionally, it identifies a recommended base of standards and de facto standards in the form of a grid Information and Communication Technology (ICT) profile, taking into account the requirements for interoperability in the ICT domain. The ETSI TR [9] considers barriers to interoperability of grid technology from the perspective of gaps in existing standards. Five areas are covered in detail: architecture, service level agreements, charging, security, and service discovery. The ETSI TR [10] describes a grid testing framework based on existing testing and validation methodologies, best practices and tools used, in the IT and Telecom sectors to obtain ICT Interoperability. It lists and

3 compiles existing grid interoperability solutions including interoperability events, state of the art papers, guidelines, interoperability profiles, reference implementations, use cases, test suites, test beds, testing tools, and open source developments. For the past few years, ETSI TC Grid has been developing approaches to improve the interoperability of grid and cloud technologies as well as enabling their use and integration into future NGN solutions. This year, TC Grid has announced a back-to-back Plugtests and workshop event about Grids, Clouds & Service Infrastructures to assess and disseminate latest solutions and trends of commercial as well as open source products in this domain [11]. III. WHY GRID/CLOUD COMPUTING IS OF INTEREST TO THE TELCO INDUSTRY A Eurescom Report on TelcoGrid about business opportunities for telecom operators in the grid market concluded that telecom operators are bound to become key players in a grid value chain as they provide connectivity and own computing resources [12]. Moreover, they have established customer relationships and accounting/billing experience, essential for business/commercial grids [13]. Reference [14] states that telecommunications services are required to support grid-enabled web service applications. Currently, no consensus exists on the specific characteristics of local, metropolitan, and wide area network services that meet an acceptable price and performance and functionality requirements of grid computing applications in the future. Several major players including British Telecommunications (BT) and Telefonica of Spain continue to make investments required to measure the market potential of grids [8]. In an article, Peter Lee [15] the CEO of DataSynapse states, "It s not news to anyone that telecommunications carriers constantly are faced with new challenges and unrelenting cost pressures. Providers are beginning to turn to a technology that has become a de facto standard at the world s largest financial institutions. Grid computing is making inroads in telco operators IT infrastructure plans helping carriers reduce costs, accelerate time-to-market, better serve growing customer bases and extend competitive advantages." In the opinion of AT&T [16], the IT industry looks for the following attributes, which are currently exhibited by telecommunications service providers, from their grid/cloud computing service providers: Enterprise sales capability, Lifecycle service and support, Reliable operations at scale, Service Level Agreements (SLA), Full enterprise solutions portfolio, Integrated hosting and network services, Vendor independence, Global footprint, Financial stability and market commitment. Deutsche Telekom Laboratories has recently spun off its own open-source cloud-computing start-up called Zimory [17]. Zimory aims to help bring the benefits of cloud computing to private enterprises. The Zimory public cloud for sellers aggregates available server computing capacity from around the world and makes it available through an Internet trading platform. France Telecom in collaboration with HP has developed a system based on grid computing that enables telecommunications carriers to optimize the usage of their IT resources [18]. This system has been tested by an automated allocation of service loads among several servers located in Paris, Tokyo, and Kawasaki in order to check if the overall system is able to handle loads that are beyond the capacity of conventional systems. BT is currently launching its Virtual Data Centre (VDC) service that aims to support large business and public sector organisations to succeed in the current economy and prepare for the future. It provides a dynamic and virtualised infrastructure platform that enables them to consume their IT and networking infrastructure as a service and forms the base for future cloud services [19]. In addition, BT has announced that it is developing Software as a Service (SaaS) for business customers. Open Telefónica is an initiative of the Strategy Unit of the Telefónica Corporate Centre to coordinate innovation activities of the different Telefónica companies with the aim of creating an ecosystem that allows customers and small developers to offer services over Telecom capabilities (X as a Service). Within the Telefónica Group, some initiatives have already been started. For example, Telefónica España s OpenMovilforum [20] and O 2 UK s Litmus [21] offer Mobile capabilities, i.e., APIs to access Mobile services. Telefónica I+D s Morfeo [22] open source community in cooperation with several partners develops different software platforms for web application development and integration (EzWeb, FAST, MyMobileWeb), collaborative development tools (Vulcano Forge) and cloud middlewares (OpenNebula, Nephele). IV. OPTIONS FOR COMBINING GRID/CLOUD AND NGN To combine NGN and grids or clouds, a number of possible architectural scenarios are under investigation [23]. These scenarios can be grouped into two main types. First, using the NGN to deliver grid or cloud services and, second, using grid or cloud services to support the NGN. Discussions between grid and NGN experts in ETSI have been initiated on the following scenarios: Grid-enabled NGN application, NGN subsystems offering grid services, Grid technology for implementing NGN functionality, Combining grid and networking resources in a new architecture.

4 A. Grid-enabled NGN application In terms of their network requirements, grid applications are highly diverse. In general, grid applications make significant use of computational and storage resources. In addition, these applications are connectivity intensive and require coordination between multiple activities. Grid applications may be either session-based (e.g., interactive computation steering, low-latency computations, real-time scene rendering in an online game) or non-sessionbased (e.g., complex workflow execution, batch processing, data movement and staging). However, they are all characterized by the need for network connectivity and, therefore, it should be possible to regard them as NGN applications. They may have specialized network requirements such as guaranteed low latency or high throughput data transfer, either sustained or intermittent. The ability for an application to interact with NGN to request the network characteristics it requires is desirable. In the first scenario, as depicted in Figure 1, a grid/cloud enabled NGN applications runs within an NGN grid application server. The NGN is already designed to support a wide range of application servers and via standard interfaces possibly other types developed by other groups. 2) Offering grid services towards end-user applications providing them with resources managed by the GSS in the same way that non-ims IPTV services are offered. 3) Dedicated grid-enabled functions of other subsystems may use the GSS. This option will require the current NGN service layer subsystems to be enhanced to support grid services in the same way that IP Multimedia Subsystem (IMS) based IPTV services are offered. C. Grid technology for implementing NGN functionality, In the third scenario, as depicted in Figure 3, logical NGN functions, or entire NGN subsystems are realized using grid services. In this scenario, logical NGN functions, from the transport stratum, like the Network Attachment Subsystem (NASS) or the Resource and Admission Control Subsystem (RACS), up to the services stratum and the applications, are implemented or realized using grid technology. This would allow the optimization of the resources used by these functions and lead to a more flexible scalability of the NGN subsystems. This is only an implementation variant but would need an additional interface to a grid service entity to manage and control grid resources. However, the NGN architecture, or some of its components, may also need to be adapted if an improvement of functionality or usability can be achieved by usage of grid infrastructures, e.g., of a peer-to-peer system on Call Session Control Function (CSCF) and Home Subscriber Server HSS. Figure 1. Grid-enabled NGN application B. NGN subsystems offering grid services In the second scenario (Figure 2), a new NGN subsystem is added to the NGN service layer to support the provision of grid services. This grid services subsystem (GSS) would give access to virtualized grid resources through a new service interface and could provide, for example, core grid services. Such an integration of a GSS would enable three different use case options: 1) Offering grid services to grid-enabled applications as defined in the first scenario. Similarly to Scenario 1, it needs to be investigated if there is any impact on already defined NGN reference points (in this option, it is between GSS and application server) and which type of NGN application server (Type 1 and/or 2) would be appropriate. Figure 2. NGN subsystems offering grid services D. Combining grid and networking resources in a new architecture Finally, in the fourth scenario, a separate Resource Management grid service manages shared resources such as computing power, network and storage. This enables the assignment of these resources to the grid, the cloud, or the NGN in a flexible, generic way. The usage of resources like computing power, network and storage resources are orthogonal to the NGN architecture

5 specifying logical functions. This allows defining or adopting standards for the management of the above resources independently of NGN details. As such resources will be needed both to run NGN-control, and as resources allocated to subscribers by NGN-control, this should finally lead to a combined architecture, allowing the assignment of all execution, storage and networking resources in a flexible, generic way. Figure 4 shows the highest level structure of such an architecture. The layer on top of the resources, called "Resource Management", manages all the available resources towards the specific services, i.e., NGN or grid services. V. IMPLICATIONS OF COMBINING GRID/CLOUD AND NGN This section identifies a number of issues, identified above, which will need to be addressed for a successful combination of grid or cloud computing with the NGN. Monitoring Brokering AAA Advance Reservation Scheduling Transport Management Data Sharing and Management Policy A. Using the NGN to deliver grid/cloud services The NGN can fulfil a number of the requirements of a commercial grid or cloud computing service, for example, by providing IP connectivity, authentication and authorization, security, transport Quality of Service (QoS) and charging. It needs to be investigated which requirements the NGN must meet in a commercial environment. The impact, if there is any, on already defined NGN reference points has to be analyzed. Also, in Scenario 2 Option 1, it has to be determined which type of NGN application server (Type 1 and/or 2) [24] is appropriate. In Option 2 and 3, grid and cloud services can be integrated into the NGN architecture in the same way as IPTV has been integrated [25]. For example, they could be integrated as a stand-alone subsystem or by using the services of an enhanced IMS. Figure 3. Grid technology for implementing NGN functionality There is a need to study which of the core and user focused grid and cloud service requirements can be provided by an NGN with integrated grid or cloud services. The ten most common requirements of grid (OGSA) services are: Discovery, Metering and Accounting Figure 4. Combining grid and networking resources in a new architecture B. Using grid/cloud services to support the NGN Grid environments as well as some cloud infrastructures are based on the Service Oriented Architecture (SOA). The NGN has a functional architecture. Their combination would be facilitated if the NGN would evolve to the SOA. It should be noted that the NGN management plane already has a SOA. However, the usage of resources like computing power, network and storage resources are orthogonal to the NGN architecture specifying logical functions. This allows defining or adopting standards for the management of the above resources independently of NGN standards. Thus, it may be possible to exploit grid/cloud, computing without impacting the NGN architecture [26]. As such resources will be needed both to run NGN-control, and as resources allocated to subscribers by NGN-control, this should finally lead to a combined architecture, allowing the assignment of all execution, storage and networking resources in a flexible, generic way. A deeper insight of the work described here is given in the article referenced in [27]. ACKNOWLEDGMENT The authors, all members of ETSI Specialist Taskforce 331, would like to thank to the members of ETSI Technical Committee GRID (TC GRID) for the guidance and support. We would also like to thank the European Commission for the funding provided under the contract SA/ETSI/ENTR/000/ which has made this paper possible. The work carried out here is co-financed by the

6 EC/EFTA in response to the EC s ICT Standardization Work Programme. ABOUT ETSI ETSI produces globally-applicable standards for Information and Communications Technologies (ICT), including fixed, mobile, radio, converged, broadcast and internet technologies and is officially recognized by the European Commission as a European Standards Organization. ETSI is a not-for-profit organization whose 700 ETSI member organizations benefit from direct participation and are drawn from 60 countries worldwide ( ETSI's Technical Committee TC GRID has the goal of addressing issues associated with the convergence between Information Technology and Telecommunications. The focus is on scenarios where connectivity goes beyond the local network. This includes not only grid computing but also the emerging commercial trend towards cloud computing which places particular emphasis on ubiquitous network access to scalable computing and storage resources. REFERENCES [1] Treadwell, J. (ed): Open Grid Services Architecture Glossary of Terms Version 1.5. GFD-I.081. Open Grid Forum (2006). [Online: fetched on ] [2] EGEE Enabling Grids for E-sciencE [Online: fetched on ] [3] EU 6 th Framework Programme (FP6) Projects in Grid Technologies [Online: fetched on ] [4] Amazon Web Services [Online: fetched on ] [5] McKinsey & Co. Report: Clearing the Air on Cloud Computing [Online: fetched on ] [6] Foster, I., Zhao, Y., Raicu, I., Lu, S.: Cloud Computing and Grid Computing 360-Degree Compared Grid Computing Environments Workshop, GCE '08 (2008). [7] IBM Cloud Computing [Online: fetched on ] [8] ETSI TR V1.1.1: GRID; Study of ICT Grid interoperability gaps; Part 1: Inventory of ICT Stakeholders. European Telecommunications Standards Institute (ETSI), Sophia-Antipolis, France (2008) [9] ETSI TR V1.1.1: GRID; Study of ICT Grid interoperability gaps; Part 2: List of identified Gaps. European Telecommunications Standards Institute (ETSI), Sophia-Antipolis, France (2008) [10] ETSI TR V1.1.1: GRID; ICT Grid Interoperability Testing Frameworkand survey of existing ICT Grid interoperability solutions. European Telecommunications Standards Institute (ETSI), Sophia- Antipolis, France (to be published) [11] ETSI Plutests Interop Events: Grids, Clouds & Service Infrastructures [Online: fetched on ] [12] EURESCOM: TelCo Grid Business Opportunities for Telecom Operators in the Grid market, EURESCOM (2004, not publicly available) [Online: series/p1349 fetched on ] [13] EURESCOM New Eurescom projects & studies [Online: dies.asp fetched on ] [14] Grid computing - a vertical market perspective , The insight research cooperation, Boonton, New Jersey, USA, 2006 [Online: fetched on ] [15] Lee, P.: Making the Impossible, Possible: The Impact of Grid Computing in Telecommunications, xchange magazine (2006) [Online: fetched on ] [16] 10 Reasons Why Telcos Will Dominate Enterprise Cloud Computing, On-Demand Enterprise [Online te_enterprise_cloud_computing_ html fetched on ] [17] Zimory: Deutsche Telekom Spinoff First to Launch Global Marketplace for Cloud Resources (2009) [Online: fetched on ] [18] France Telecom and HP: Together, we can do more, CIO White Paper (2009) [Online: fetched on ] [19] Q&A: BT Business head of SaaS, Chris Lindsay [Online: fetched on ] [20] Open movil forum [Online: fetched on ] [21] O 2 Litmus Beta [Online: fetched on ] [22] MORFEO Project [Online: fetched on ] [23] ETSI TR V1.1.1: "GRID; Grid Services and Telecom Networks; Architectural Options " European Telecommunications Standards Institute (ETSI), Sophia-Antipolis, France (2009) [24] ETSI TS : "Digital cellular telecommunications system (Phase 2+); Universal Mobile Telecommunications System (UMTS); LTE; Network architecture (3GPP TS )". European Telecommunications Standards Institute (ETSI), Sophia-Antipolis, France (2009) [25] ETSI TS : "Telecommunications and Internet converged Services and Protocols for Advanced Networking (TISPAN); Service Layer Requirements to integrate NGN services and IPTV". European Telecommunications Standards Institute (ETSI), Sophia-Antipolis, France (2007) [26] ETSI ES : "Telecommunications and Internet converged Services and Protocols for Advanced Networking (TISPAN); NGN Functional Architecture". European Telecommunications Standards Institute (ETSI), Sophia-Antipolis, France (2007) [27] Rings, T. et al.: Grid and cloud computing: opportunities for integration with the Next Generation Network. Journal of Grid Computing: Special Issue on Grid Interoperability, JOGC, in press