一种超节点P2P网络中基于语义的协同缓存管理机制

超节点P2P（Super-peerP2P）结合了P2P结构和C／S结构的优点，是当前应用最广的一类P2P系统。在超节点P2P网络中，文件访问是最基本的操作，往往使用缓存技术来提高其操作效率。目前大多数超节点P2P网络使用传统的“尽力而为”的缓存机制，该机制没有区分超节点网络中不同节点对资源的需求及关注程度不同，导致偶尔访问的对象替换经常访问的对象。针对“尽力而为”缓存机制的不足，本文提出一种基于语义信息的协同缓存管理机制SCOCM（Semantic-based Cooperative Cache Management mechanism for super-peer net works），应用已经请求对象的语义信息主动地选择对象放置缓存，以兴趣度的远近驱逐缓存内的对象，减少缓存对象的替换，使得每个缓存中缓存的对象之间尽可能地保持语义联系。实验结果表明，基于语义信息的协同缓存管理机制与LRU相比可大大降低缓存的替换率，提高缓存的存取效率，命中率也较高。

A Semantic-based Cooperative Cache Management Mechanism for Super-peer Networks

As one of the most important kinds of peer-to-peer(P2P) systems, super-peer networks takes the advantages of P2P model and C/S (client/server) model for its architecture which combined the two models. Most of the super- peer systems apply the caching approach to improve the access performance. Generally speaking, super-peer networks use the best-effort caching mechanism. However, the best-effort caching mechanism does not distinguish the require- ments and attentions from different peers so that the objects accessed occasionally may replace those often accessed in the cache. This paper proposes a semantic-based cooperative cache management mechanism for super-peer networks called SCOCM. Applying the semantic information of the objects already in the cache, SCOCM chooses the potential accessed objects forwardly and use them to replace the "remote" objects in the caches according to the peer interests to reduce the replacement frequencies. SCOCM tries to keep the potential semantic relations among objects in each cache in the peer clusters so the queries can be processed cooperatively in the clusters. The experiment implies that SCOCM can improve efficiency of searching and the hit ratios of the caches.