Research On Cache Replacement Policy Of CDN

With the development of Internet technologies and Web services, Web traffic still increases exponentially. The architecture of Client/server results in severe network bandwidth scarcity, network congestion and server overloading, which are the key issues affecting the continuous growing of the Internet. Users have to stand with longer request delays.CDN (Content Delivery Network) is the main and effective technique to solve these issues. CDN can push the content and service to the edge of Intenet by utilizing an intermediate layer with a set of service nodes that have been predeployed to store the contents of the sites, so that the distributed content is on the network edge which is closer to the user, thus speeding up the network, reducing duplication of redundant data transmission and transmitting the repeated data to users by local cache server. It also alleviates the burden on the sevrer and the network saves bandwidth resources.A conclusive reason for the success of CDN is that users can read the contents that they need from the nearest CDN cache server. One of the important factors of improving the efficiency of CDN is to increase the cache hit rate. Due to the limited capacity of the cache server, it is not realistic to cache all the contents. Therefore, the cache can only replace the contents of low access rate and store the contents of high possibility of repeated visit by some strategy. Replacement policy is an important part of the cache management. Replacement policy is to design a strategy to remove the contents periodically popular that users are not interested in so as to retain sufficient space for more popular contents in order to get a higher hit rate.The effectiveness of the replacement policy is closely related to the features of the web business. If the cache replacement policy does not reflect the request model, the cache server will not be able to receive a high request hit rate.By analyzing the documents of web trace, we found the uer's request characteristics as below:1. The contents that the users request appear to be local in time: the users tend to request the document again that they recently requested, which is that the users have the same interest to requesting the contents.2. The users always tend to request the smaller contents: the users are prone to request the smaller ones when they request the web contents.Finally, based on the analysis of Web trace, we found the characters of the Web content request. According to these characters we got the probability of the content request by the user to be requested again. We called it reusable ratio. Through this ratio we presented a cache replacement policy of the Content Delivery Network (CDN) to improve the hit rate of the cache. Show the superiority of the replacement policy over the existing policy via numerical simulation computation.