Search Result Caching in Peer-to-Peer Information Retrieval Networks

Abstract

For peer-to-peer web search engines it is important to quickly process queries and return search results. How to keep the perceived latency low is an open challenge. In this paper we explore the solution potential of search result caching in large-scale peer-to-peer information retrieval networks by simulating such networks with increasing levels of realism. We nd that a small bounded cache oers performance comparable to an unbounded cache. Furthermore, we explore partially centralised and fully distributed scenarios, and nd that in the most realistic distributed case caching can reduce the query load by thirty-three percent. With optimisations this can be boosted to nearly seventy percent. IntroductionIn peer-to-peer information retrieval a network of peers provide a search service collaboratively. We dene a peer as a computer system connected to the Internet.The term peer refers to the fact that in a peer-to-peer system all peers are considered equal and can both supply and consume resources. In a peer-to-peer network each additional peer adds extra processing capacity and bandwidth in contrast with typical client/server search systems where each additional client puts extra strain on the server. When such a peer-to-peer network has good load balancing properties it can scale up to handle millions of simultaneous peers.However, the performance is strongly aected by how well it can deal with the continuous rapid joining and departing of peers which is called churn.We study peer-to-peer information retrieval systems where each peer contains, and maintains an index over, a subset of all the documents in the system.Since presumably relevant search results can be located at any peer it is dicult to route a query to the right one. This problem is commonly approached using dierent network topologies and replication of index data [1,2,3].In this paper we explore search result caching. We assume that for each query there is a peer that can provide a set of original search results. If this query is posed often that peer would cripple under the demand for providing this set over and over again. Hence, we propose that each peer that obtains search results for a particular query caches those results. The eect is that search results for popular queries can be obtained from many peers: high availability, and the load on the peer that provided the original results is reduced: load balancing.