A Two-level Overlay Approach with Intelligent SwitchAnshi LiangYe ZhouA Two-Level Overlay Approach with Intelligent Switch Anshi Liang Ye Zhou {lasnake, [email protected]} Abstract This paper describes a new approach in overlay network— a two level overlay structure. In particular, it describes the usage of this new approach in content delivery. In the traditional overlay network, nodes are placed at the edge points of LAN. This kind of approach is simple but lacks the ability of QoS. In this paper, we proposed to add one more level of overlay nodes in the middle of network, which can help us to achieve QoS content delivery. In addition, we proposed to equip our overlay architecture with Alteon+ISD[1], a unique combination with switching and computational power. We address the benefits of such architecture and analyze the efficiency in aggregating traffics. We propose a business model and plan to extend our 2-level overlay idea to other applications such as storage network and wireless network. 1. Introduction The original Internet architecture was designed to provide stateless, unicast point-to-point communication between fixed locations. The simplicity of this stateless communication abstraction contributed greatly to the scalability and efficiency of the Internet. The traditional networking research is based on the end-to-end principle[15]: --The function in question can completely and correctly be implemented only with the knowledge and help of the application standing at the endpoints of the communications system. Therefore, providing that questioned function as a feature of the communications system itself is not possible The end-to -end approach has a number of advantages: 1. the complexity of the core network is reduced; 2. a new application can be added without having to change the core of the network; 3. applications do not have to rely on the successful implementation and operation of application-specific services in the network. As a consequence, much of the Internet’s design was based on the end-to-end arguments. However, there have been significant changes in recent years compared to the early stage of Internet[14]. From the application point of view, there are more demanding applications, ISP service differentiation, more security issues and less sophisticated users. The traditional end-to-end network design approach cannot meet these challenges. On the other hand, from the hardware point of view, there are more powerful computer boxes and more sophisticated routers/switches, so now we do not need to limit the functionalities at end nodes; we can have more functionality in the middle of network, besides the general data transfer service. Overlay network is one of the attempts to change/modify the current end-to -end network design. Overlay network adds an additional layer of indirection/visualization to achieve certain functionalities in the underlying network. Akamai[6] and i3[4][5]are examples. However, the current approaches of overlay network cannot provide QoS content delivery. In this paper, we propose to have a two-level overlay network to achieve the goal of QoS content delivery. In this approach, the inner level nodes are placed on the Metropolitan Area Network (MAN) for QoS channel and service negotiation, traffic aggregation, etc. The outer level nodes are placed at the edge of Local Area Network (LAN) for tasks such as content delivery, multicast, admission control, replicated audio/video streaming, etc. In the latter part of this paper, we show that this two level overlay approach can significantly reduce the traffic for content delivery and the number of route setups; it also increases the efficiency of capacity usage. The overlay nodes of our approach need strong switching and computational power. We propose to use the Nortel Alteon+ISD combination[1] (programmable gateway) to serve as our overlay nodes. (A brief structure description can be found in the appendix.) This unique combination enables the functionalities of our two-level overlay network. The usage of this two-level overlay network can be extended to other applications such as storage network and wireless network. In the next section, we give a brief overview of current approaches in overlay network, specified in content delivery. In section 3, we will talk about the design of our two-level overlay architecture. In section 4, we will give a theoretical analysis to our overlay model and propose a business model. Section 5 includes our plan for verification. We conclude in section 6 with a brief discussion of open problems and future work. 2. Current Approaches on Overlay Network 2.1 Akamai Akamai presents an overlay structure in web content delivery. (See Figure 2.1) It has 13000 web cache servers all over the world, these cache servers are serving as the nodes of this overlay structure. They are placed at the edge point of local area networks. In this model, web sites do not send contents to end users upon request. Instead, it sends contents to the web cache servers. The web cache servers assemble, replicate and then multicast the contents to end users. The congestion between web cache servers and end users is minimal because the web cache servers are placed close to end users. The traffic of content delivery is substantially reduced because web sites only send copies to the web cache servers in stead of sending individual copies to end users. The web cache servers can also perform tasks such as load balancing and content storage. ｕｓｅｒ ｕｓｅｒｕｓｅｒｕｓｅｒｕｓｅｒｕｓｅｒＣｏｎｔｅｎｔ Ｐｒｏｖｉｄｅｒ ｕｓｅｒ ｕｓｅｒＭＡＮｕｓｅｒｕｓｅｒｕｓｅｒｕｓｅｒｕｓｅｒｕｓｅｒｕｓｅｒｕｓｅｒｕｓｅｒｕｓｅｒｕｓｅｒｕｓｅｒｕｓｅｒｕｓｅｒｕｓｅｒｕｓｅｒｕｓｅｒｕｓｅｒｕｓｅｒｕｓｅｒｕｓｅｒｕｓｅｒｕｓｅｒｕｓｅｒｕｓｅｒｕｓｅｒＣｏｎｔｅｎｔ Ｐｒｏｖｉｄｅｒ Figure 2.1 Figure 2.2Akamai provides good service for web content delivery, but it cannot deal with content delivery with QoS requirements or other real time traffics. Currently Akamai is using best-effort method to deliver contents from web site’s own server to web cache servers. A natural thinking is to establish QoS
View Full Document
Unlocking...