A Request-based Dynamic Leader Adjustment Method In CDN

With the rapid development of internet technology and the continued diversity of network services, the old client-server model can't satisfy the growing demand for network. CDN,caching replication technology has become the key to solve these problems. And because of the existence of documents copies, Consistency strategy must be deployed to ensure the data is the latest for the users. Co-sponsoed by the union lab of Beijing ChinaCache Communication Co.,Ltd and Huazhong University of Science and Technology—"Research of the Pull/Push strategy in Content Delivery Network", in this thesis, we have launched in-depth study of how to keep consistency of CDN network with efficient strategies, also, propose a request-based dynamic leader adjustment strategy.This thesis first expounded the definition of Content Consistency, and noted that the consistency strategy exist in single proxy server. Then we compared the advantages and disadvantages of the various strategies, and the reference to CDN. Considering the characteristics of CDN, it is important to set "Lease" to ensure consistency, also, analysis of all kinds of especial Lease, including the cooperation of all Leases, and the setting of key parameters. There is a detailed introduction of the process of Lease strategy, and how it can be expanded to a multi-level network and adaptively establish tree-based network. It also described how to recover from the sudden failure.In view of the current Leader selection strategy, A request-based dynamic leader adjustment strategy was presented in this thesis. To improve the cache hit rate of bytes and reduce the user's response time, the proxy server which receiving more requests as the leader of the region can reduce the update data and bandwidth consumption by exchanging small amount of information,because the proxy server which receiving more request in the follow-up will receive more requests.Trace-based simulation had proved that this strategy can reduce response time 4% -20% when the bandwidth consumption increase is less than 5%. Trace-based simulation had proved the request-based dynamic leader strategy worked well. The follow-up work can be done to set up adaptive threshold values to enhance efficiency.