Contention-Based Performance Evaluation of Multidimensional Range Search in Peer-to-peer Networks

Performance evaluation of peer-to-peer search techniques has been based on simple performance metrics, such as message hop counts and total network traffic, mostly disregarding their inherent concurrent nature, where contention may arise. This paper is concerned with the effect of contention in complex P2P network search, focusing on techniques for multidimensional range search. We evaluate peer-to-peer networks derived from recently proposed works, introducing two novel metrics related to concurrency and contention, namely responsiveness and throughput. Our results highlight the impact of contention on these networks, and demonstrate that some studied networks do not scale in the presence of contention. Also, our results indicate that certain network properties believed to be desirable (e.g. uniform data distribution or peer accesses) may not be as critical as previously believed.