Preserving Consistency of Dynamic Data in Peer-Based Caching Systems

摘要

On-line decision making often involves significant amount of time-varying data, which can be regarded as dynamic data. Maintaining dynamic data consistency in existing P2P systems is often inefficient. Most of the previous P2P research, works are predominantly remaining at, the static, files management. In this paper, we focus on the problem of maintaining consistency of dynamic data items in an overlay network of cooperating peers. We present CC-Buddy, an adaptive framework for efficiently disseminating dynamic data in P2P environments. Our approach combines application-level multicasting and demand-driven dissemination filtering techniques. We evaluate CC-Buddy with a prototype system built on top of BestPeer infrastructure. We examine our approach using real-world traces of dynamically changing data items. The results show our techniques can efficiently satisfy the consistency requirements by different users.