Focus On The Resource Allocation Among Large Scale Overlay Networks

As the bridge between emerging applications and future IP networks, Overlay Networks have experienced a significant increase recently. In essence, Overlay Networks guarantee their QoS by constructing a virtual layer on the top of IP networks and optimizing their resources according to the different applications. So it is significant and important to allocate the limited network resources among Overlay Networks effectively.This dissertation aims to provide an optimal resource allocation framework for Overlay Networks to delivery content effectively, so as to service end users with high QoS. To this end, we try to deal with the following key challenges:(1) how to select the proper locations to deploy service nodes and assign the limited network resources to these nodes appropriately to optimum static resource among Overlay Networks; (2) how to deliver service content from source service nodes to end users quickly and effectively in specific applications, which includes two processes:allocating the service resource of source service nodes and edge service nodes. The former aims to allocate the network resource among different virtual links appropriately to achieve high effectiveness of the content delivery and minimize the delay of request response when the service content is transferred from the source service nodes to the edge service nodes, and the latter tries to apply appropriate resource allocation strategies to provide guaranteed QoS for more end users under the worst Internet conditions at Internet edge.Facing the above challenges, this dissertation makes efforts and contributions in the following aspects.(1) According to the natures that Overlay Networks need to select service nodes actively, we proposed a network coordinates-based server placement framework. In this framework, by taking advantage of the network coordinates-based Internet measurement, we get the complete Internet information with a little measurement overhead, and then we applied a modified hierarchical clustering to get optimal deployment strategy in global Internet, which can overcome the shortcomings of the traditional server placement models in which the service nodes must be chosen from these candidates.(2) We presented a server placement model, which is suitable for different application systems and different Internet conditions. Our model takes into account the factors of the deployment cost, user experience and Peer contribution rate, and then we integrated these factors into our optimal goal to adapt to different Internet conditions and different applications.(3) We proposed a resource allocation strategy for the content delivering process from source service nodes to edge service nodes, which was based on the relationship between the flow information of the physical networks and the application layer flow of Overlay Networks. To apply this strategy, we first collected the physical network flow information and the current resource request information, and then allocated the network resource among different virtual layer links to optimize the service performance of Overlay Networks.(4) We presented a dynamic resource allocation strategy and an adaptive resource allocation algorithm for edge service nodes to deal with the universal low quality of applications under the worst Internet conditions. Based on our P2P scalability model, and on the fact that there are two kinds of Peers (random Peers and On-device Peers) in P2P systems and On-device Peers always exist, we introduced the controllability parameter of P2P systems to optimize the resource allocation among different end users.