How caching queries at client-peers affects the loads of super-peer P2P systems

Super-peer P2P systems strike a balance between searching efficiency in centralized P2P systems and the autonomy, load balancing and robustness provided by pure P2P systems. A super-peer is a node in a super-peer P2P system that maintains the central index for the information shared by a set of peers within the same cluster. The central index handles the searching request on behalf of the connecting set of peers and also passes on the request to neighboring super-peers in order to access additional indices and peers. In this paper, we study the behavior of query answering in super-peer P2P systems with the aim of understanding the issues and tradeoffs in designing a scalable super-peer system. We focus on where to post queries in order to retrieve the result and investigate the implications for three different architectures: caching queries at the peer; caching only at the super-peer; and an ordinary P2P system without any caching facilities. We are adopting the existing equation on measuring the network cost for query answering in super-peer systems. In addition, we are adapting the same equation for super-peer system with caching facilities while answering their queries. Using these equations, the cost of query processing for these architectures is compared. The paper discusses the tradeoffs between architectures with respect to caching, highlights the effect of key parameter values on system performance, and ends by considering whether certain knowledge domains are more appropriate for particular architectures.