A Cache Replacement Policy Based On Content-Centric Networking

With the rapid development of the Internet, the demand for network content is increasing. The current TCP/IP network architecture has the increasing serious problem in network control,resource allocation and so on. The focus of Internet communication has been transferred from the host-to-host to how fast access to the requested content from the Internet. For this reason, the domestic and foreign scholars began to research the next generation network architecture, and launch a lot of researches on related projects,promoting the development of the next generation network, which is of epoch-making significance. This paper studies a new network architecture which is called "content center networking",CCN abandoned the traditional network's core communication with the host address, but turned to the named-content as its thought. In order to adapt to the development of future network, CCN builds a new architecture and communication mechanism.Cache replacement policy which is related to the overall performance of the network is also a key part of the CCN. Often used cache replacement policies in CCN are LRU and LFU and their improvement policy. These policy can improve the efficiency of network communication, but also simple and easy to implement. However, they did not fully consider the dynamic characteristics of CCN, so they have a large degree of inadaptability.In this thesis,we propose a cache replacement policy based classification of content popularity. The policy combines the method of all-gram and R-value to operate contents'feature extraction and classification according to the way of content naming and the characteristics of uniqueness of names, so the node can manage the cache on classification.In addition to this, the calculation of the popularity of categories in each node with an exponential weighted moving average. Giving different weight by the time distance and the number of times to be accessed in predetermined time each category. In the cache replacement process, replace the content item which in the category with the lowest popularity in the node, then put new content into corresponding category according to the classification method.Finally,we set up an experimental simulation environment,with the network similar NS-3, we tested the performance of proposed cache replacement policy, and compared with other commonly used cache replacement policies. We evaluate the performance of average cache hit ratio and average server load according to different cache memory size and different number of stub domain. Experiments proved that the policy we proposed can well adapt the demand for content caching policy in dynamic network while improving the cache efficiency of the whole network and have certain advantages in performance compared to other policies.