Network Working Group Z. Zhang Internet-Draft Z. Sun Intended status: Standards Track J. Su Expires: May 14, 2010 NUDT November 10, 2009 Problem Statement of Core Network Overlay draft-zhang-cno-problem-statement-00.txt Abstract With the convergence of computing, storage and network, especially the maturity of data center technology, it becomes a trend that ISP would like to deploy the computing and storage resources in the network, so that the optimized resource utilization as well as the value-added service can be provided. The data centers deployed in the network can provide the runtime environment and resource to optimize ICP services with various edge service proxies, such as CDN servers and distributed Trackers. Meanwhile, based on data centers ISP can provide all kinds of value-added services in the IP based network. CNO (Core Network Overlay)is an ISP-controlled overlay network constructed on top of the core network which is proposed as a solution to large-scale data distribution in the Internet. It can provide the transfer service for data distribution and switching in the core network for the ICP and its edge service proxies. Because the CNO nodes are trustworthy (controlled by ISP), stable and have a simple economic model (without any participation of end users and ICPs), various overlay based data distribution methods are easier to be applied and standardized with CNO. This draft presents the definition of the conceptions and terms about CNO. Status of this Memo This Internet-Draft is submitted to IETF in full conformance with the provisions of BCP 78 and BCP 79. Internet-Drafts are working documents of the Internet Engineering Task Force (IETF), its areas, and its working groups. Note that other groups may also distribute working documents as Internet- Drafts. Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress." The list of current Internet-Drafts can be accessed at Zhang, et al. Expires May 14, 2010 [Page 1] Internet-Draft Problem Statement of Core Network Overlay November 2009 http://www.ietf.org/ietf/1id-abstracts.txt. The list of Internet-Draft Shadow Directories can be accessed at http://www.ietf.org/shadow.html. This Internet-Draft will expire on May 14, 2010. Copyright Notice Copyright (c) 2009 IETF Trust and the persons identified as the document authors. All rights reserved. This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (http://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the BSD License. Zhang, et al. Expires May 14, 2010 [Page 2] Internet-Draft Problem Statement of Core Network Overlay November 2009 Table of Contents 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 4 2. Definitions . . . . . . . . . . . . . . . . . . . . . . . . . 5 3. Design Issues . . . . . . . . . . . . . . . . . . . . . . . . 5 3.1. Reference Model . . . . . . . . . . . . . . . . . . . . . 5 3.2. CNO's Impact on Each Party . . . . . . . . . . . . . . . . 6 3.2.1. Impact on ISP . . . . . . . . . . . . . . . . . . . . 6 3.2.2. Impact on ICP . . . . . . . . . . . . . . . . . . . . 7 3.2.3. Impact on End-User . . . . . . . . . . . . . . . . . . 7 3.3. Scope of Work . . . . . . . . . . . . . . . . . . . . . . 7 3.4. Influences on Internet Evolution . . . . . . . . . . . . . 7 4. Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 4.1. Interest Relationship among ISPs . . . . . . . . . . . . . 7 4.2. Relationship with End-to-End . . . . . . . . . . . . . . . 8 5. Related Researches . . . . . . . . . . . . . . . . . . . . . . 8 5.1. Tussle . . . . . . . . . . . . . . . . . . . . . . . . . . 8 5.2. ISP Post Offices . . . . . . . . . . . . . . . . . . . . . 8 5.3. JUNOS . . . . . . . . . . . . . . . . . . . . . . . . . . 8 5.4. Programmable Overlay Router . . . . . . . . . . . . . . . 9 5.5. Data Center . . . . . . . . . . . . . . . . . . . . . . . 9 6. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 9 7. Security Considerations . . . . . . . . . . . . . . . . . . . 9 8. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 9 9. Informative References . . . . . . . . . . . . . . . . . . . . 9 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 10 Zhang, et al. Expires May 14, 2010 [Page 3] Internet-Draft Problem Statement of Core Network Overlay November 2009 1. Introduction Internet video reached 31 percent by the end of 2008 (not including the amount of video exchanged through P2P file sharing) and Cisco forecasts that the sum of all forms of video will account for close to 90 percent of consumer traffic by 2012 [Cisco]. Including video application, large-scale data distribution becomes more popular in Internet. Although P2P technique could effectively support large- scale file sharing and streaming media distribution applications in Internet, it faces with the challenges of improvement in utility efficiency of network resources and avoid harm to ISP's benefit. Some prior work like P4P [P4P] and AUTO [RFC5693] utilize the network information provided by ISPs to select optimal peers and reduce cross-domain traffic, but the instability of peer nodes and lack of pricing model continues to hinders P2P from becoming a standard transfer service to users. The mechanism which efficiently supports large-scale data distribution SHOULD effectively resolve the conflict of interest among users, ISP and ICP. And it should have a relatively stable service capacity. With the convergence of computing, storage and network, especially the maturity of data center technology, it becomes a trend that ISP makes use of its computing and storage resources to optimize resource utilization and provide value-added service by deploying service infrastructure in the network. We define CNO (Core Network Overlay) mainly based on the following observations. (1) End-To-End data distribution in Internet traverses edge network where end users locate and core network operated by ISPs, while ISPs are only concerned with the problem of core network's resource optimization. (2) In order to provide users with better service quality and enhance their competitiveness, ICP is inclined to deploy service proxies (ESP: Edge Service Proxy) (e.g., CDN servers) at the network edge adjacent to users. (3) The network layer technology itself (e.g., IP Multicast) is difficult to provide data distribution services for the whole Internet. Based on the above observation, we claim that (1) ISP has an incentives to deploy data centers in the network and build runtime environment for the ICP's ESP Because ICP has the willingness to deploy more ESP based on ISP's resource. (2) The deployment of ESP partitions the data distribution in Internet into the core and edge network implementation. Data distribution in the core network is carried out among servers and ESPs of each ICP, while data distribution in the edge network is realized among ESPs and end users of each ICP. CNO's idea is to establish overlay on top of the core network through the combination of ISP's computing and storage capacity in the network. CNO provides ICPs with data switch service in the core network as well as optimizes the usage of core network Zhang, et al. Expires May 14, 2010 [Page 4] Internet-Draft Problem Statement of Core Network Overlay November 2009 resources. Under the framework of CNO-based transport service, ICP only focus on the management of the data content it owns and for the ESPs how to enhance service quality to end users at the edge network. 2. Definitions The following terms have special meaning in the definition of Core Network Overlay (CNO) problem. Core network: network which ISPs invest to construct and manage, typically including transit ISP's network and local ISP's network. Edge network: network which users invest to construct and manage, e.g., enterprise network and campus network. Overlay: Overlay refers to the network built above underlay network. CNO: overlay network built above the core IP network. NSI (Network Service Infrastructure): NSI refers to the network service infrastructure which has a certain computing and storage capacity. NSI is usually implemented as ISP's data center and can also be embedded in the service plane of routers. ESP: An ESP refers to a program which runs above NSI and provides data distribution service for edge users. It's usually deployed by ICP. OVEN (OVErlay Node): OVEN is a CNO node which runs above NSI. This program not only provides data transfer service above the core network for ESP but also implements routing of CNO network as a router. 3. Design Issues 3.1. Reference Model By the deployment of NSI, ISP can build overlay network above the core network for content distribution. Because OVEN is controlled by ISP and relatively stable, the problem of unpredictable join and exit seldom occurs. Meanwhile, since the scope of nodes is limited, it's convenient to implement routing, forwarding and other mechanisms. CNO's consumer is ICP's ESP instead of normal end users, so pricing and security issues are easier to settle. CNO's reference model is shown in Figure 1. OVEN provides ESP with Zhang, et al. Expires May 14, 2010 [Page 5] Internet-Draft Problem Statement of Core Network Overlay November 2009 the interface of data transfer service, and also realizes forwarding and routing functions of CNO. Multiple ESPs can host in one NSI and OVEN can provide different ESPs with diverse transfer services. For instance, ESP1 and ESP2 hosting on A is respectively service program for CDN and live video service program. OVEN1 offers ESP1 non real- time data transfer and offers ESP2 real-time data transfer. Because of the differences among transfer service, CNO charges diverse prices for ICPs, and chooses distinct overlay transfer mechanisms according to the transfer's demand. Based on the satisfaction of ESP's needs, CNO optimizes the resources utility of the core network. +----+ +----+ |ESP1| |ESP2| +----+ +----+ +----+ +----+ |ESP1| |ESP2| | | +----+ +----+ | | / \ | | / \ | |OVEN3 / \ | +-------------@---------- @ OVEN4 \ | / | /| \ | / | / | \|/ OVEN2 | / | OVEN1 @-------@---------+-------+ | CNO network | | | | | | | | +----+-------+---------+-----------+---+ / | | | | / / NSIA @ | | NSI D @ / / | @ NSI C / / @ NSI B / +--------------------------------------+ IP network Figure 1: CNO Reference Model 3.2. CNO's Impact on Each Party 3.2.1. Impact on ISP ISP needs to invest in the construction of NSI in the network and deploy OVEN above it. The benefits ISP obtains mainly include three aspects. First, ISP sells its resources to ESP for supporting ICP's operation. Second, ISP charges ICP for value-added services provided by CNO. Third, resource optimization of the core network is achieved by OVEN independently without the influence of end users and ICPs. Zhang, et al. Expires May 14, 2010 [Page 6] Internet-Draft Problem Statement of Core Network Overlay November 2009 3.2.2. Impact on ICP ICP purchases NSI's computing and storage resources from ISP and pays for the data transfer service. The benefits they obtain include two aspects. The first is to reduce infrastructure investment and operation costs. For instance, CDN providers simply need purchase resources of NSI instead of deploying servers globally. This makes ICP more focus on the management of content itself. The second is for ICP to quickly deploy ESP in close proximity to the end users. ESP can provide users with better quality services and enhance ICP's competitiveness. 3.2.3. Impact on End-User CNO is transparent to end users. As the ESP is deployed by ICP close to users, users will get better service. 3.3. Scope of Work We recommend that the study includes the establishment of CNO network, routing and traffic engineering. Currently, various data distribution methods based on overlay can be standardized under the framework of CNO. The communication protocol among ESPs and how to provide services for end users is not within the scope of our discussion. 3.4. Influences on Internet Evolution To some extent, CNO is a intermediate state of Internet evolution. The development and deployment of network-layer technology among NSIs, e.g. multicast, QoS guarantee, multipath and traffic engineering, is an ultimate means to solve the problem of large-scale data distribution. Besides, the evolution process that data dissemination transforms from CNO to network-layer-based technology is transparent to ESPs and end users. 4. Problems 4.1. Interest Relationship among ISPs Although the deployment of CNO isolates the interests between users and ICPs, but how to balance interest relationship among ISPs needs further study. Zhang, et al. Expires May 14, 2010 [Page 7] Internet-Draft Problem Statement of Core Network Overlay November 2009 4.2. Relationship with End-to-End The relationship between CNO and end-to-end design philosophy requires further study [End-to-End]. 5. Related Researches 5.1. Tussle We propose CNO to resolve tussle among various parties of Internet, and address the problem that the diverse mechanisms which the interests of different Internet players conflict with each other. Moreover, CNO is also in pursuit of the traditional objectives, such as the enhancement of reliability and optimization of network load. We follow the design principle [Tussle] of how to deal with tussle. First of all, the modular design is needed to the borders. CNO is introduced to separate user's application type from data distribution mechanism and QoS guarantee provided by CNO. In addition, CNO support various parties to express their wishes via choice. Users can choose whether to utilize ESP to improve user experience. ICP may choose whether to make use of the resource provided by NSI to improve the quality of service. ISP can also choose whether to deploy NSI. Likewise, there is tussle issue of service pricing in CNO. 5.2. ISP Post Offices Each domain reaches the peak of network load at different time. By the deployment of storage Post Office ISP can time-shift traffic [Hotnets08]. That is to move DT traffic from peak to non-peak time in order to reduce core network!_s bottleneck load. NSI can also adopt this mechanism to alleviate the tension of bandwidth. 5.3. JUNOS Considering how to deploy services in IP network, James Kelly [PRESTO'09.JUNOS] pointed out that JUNOS SDK tools can be utilized to provide third-party software with programmable interface of packet processing. In addition, it also supports the integration of multi- party components to form a common service platform, thus services plane can be deployed in Juniper routers. From this paper, it can be concluded that network equipment vendors have begun to consider create service function of router in which our proposed ESP of CNO can hosts. Zhang, et al. Expires May 14, 2010 [Page 8] Internet-Draft Problem Statement of Core Network Overlay November 2009 5.4. Programmable Overlay Router Bruce [PRESTO'09.POR] proposed to establish overlay with the support of programmable router provided by ISP. Through this overlay, ISP can make better use of network topology and status information to control network resources. Authors believe that ISP can provide diverse service by means of overlay that is a great power to drive ISP deploy overlay. The paper names this overlay service-routing layer and recommend the built-in function directly support the application development. In our CNO, services provided by service- routing layer can be implemented in OVEN as a common function which can be called by ESP. This can optimize each CP's service quality and improve network status. 5.5. Data Center Traditional data center is server clusters deployed by ICP to support data access requests from a large number of users. Our proposed NSI is established by ISP in the network to serve each ICP, and actually indirectly serve users under the schedule of ESP. In addition to traditional data storage capabilities, NSI also needs to provide other functions like peer selection. In respect to cost, data center located at the edge network needs to control its scale. 6. IANA Considerations This document makes no request of IANA. 7. Security Considerations The security issues brought by CNO is what we need take into consideration. 8. Acknowledgments 9. Informative References [Cisco] Cisco, "Approaching the Zettabyte Era", September 2008. [End-to-End] Salter, H., Reed, R., and D. Clark, "End-to-End Arguments In System Design", Second International Conference on Distributed Computing Systems, 1981. Zhang, et al. Expires May 14, 2010 [Page 9] Internet-Draft Problem Statement of Core Network Overlay November 2009 [Hotnets08] Laoutaris, N. and P. Rodriguez, "Good things come to Those Who (Can) Wait", Hotnets08 , August 2008. [P4P] Xie, H., Yang, Y., Krishnamurthy, A., Liu, Y., and A. Silberschatz, "P4P: Provider Portal for Applications", SIGCOMM'08 , August 2008. [PRESTO'09.JUNOS] Kelly, J., Araujo, W., and K. Banerjee, "Rapid Service Creation using the JUNOS SDK", PRESTO'09 , August 2009. [PRESTO'09.POR] Davie, B. and J. Medved, "A Programmable Overlay Router for Service Provider Innovation", PRESTO'09 , August 2009. [RFC5693] Seedorf, J. and E. Burger, "Application-Layer Traffic Optimization (ALTO) Problem Statement", RFC 5693, October 2009. [Tussle] Clark, D., Wroclawski, J., Sollins, K., and R. Braden, "Tussle in Cyberspace: Defining Tomorrow's Internet", SIGCOMM'02 , August 2002. Authors' Addresses Zi Wen. Zhang NUDT 47 Yanwachi St Changsha China Phone: +86-13975137411 Email: ziwen@nudt.edu.cn Zhi Gang. Sun NUDT 47 Yanwachi St Changsha China Phone: +86-13875903721 Email: sunzhigang@nudt.edu.cn Zhang, et al. Expires May 14, 2010 [Page 10] Internet-Draft Problem Statement of Core Network Overlay November 2009 Jin Shu. Su NUDT 47 Yanwachi St Changsha China Phone: +86-13787088961 Email: sjs@nudt.edu.cn Zhang, et al. Expires May 14, 2010 [Page 11]