Optimal layer division for low latency in DHT-based hierarchical P2P network
Jun Li,Cuilian Li,Zhaoxi Fang,Haoyun Wang,Yaohui Wu
DOI: https://doi.org/10.1002/nem.1922
2016-01-01
International Journal of Network Management
Abstract:Hierarchical architecture can significantly reduce the workload overheads and facilitate effective search in peer-to-peer P2P network by dividing the peers into two layers: a supernode layer and an ordinary node layer. System scalability is therefore improved, which is critical to the deployment of P2P applications, particularly for the P2P-based Voice over Internet Protocol systems. However, the lack of appropriate size ratio of the supernode layer to the ordinary node layer makes the system search performance far from being optimal. To address the problem, the goal of this paper is to find the optimal layer ratio, such that lookup latency is minimized, subject to the capacity constraint of supernodes. With this particular objective, this paper first formulates a search delay model for the distributed hash table DHT-based layered P2P architecture, taking account of two main sources of latency: DHT routing and queuing delay at the supernode. Then the optimal ratio between supernodes and ordinary nodes is found with the analytical model. Simulation and numerical results show that the obtained solution can improve the performance of a two-layer P2P system in terms of lookup latency. Copyright © 2016 John Wiley & Sons, Ltd.