Dynamic Load Balancing with Multiple Hash Functions in Structured P2P Systems

P2P systems can be used to form a low latency decentralized data delivery system. Structured P2P systems provide both low latency and excellent load balance with uniform query and data distributions, however, in the real application environment, requests distributions are often skewed and follow a Zipf law, some objects will become hotspots, individual nodes are easily overloaded, resulting in poor global performance and lost messages. This paper describes a novel approach with multiple hash functions to replicate the hotspots in a series of different nodes to distribute the high load evenly, and it can increase or decrease the replicas dynamically. At the same time it maintains low access latencies and good load balancing even under highly skewed demand. Results from performance evaluation demonstrate the effectiveness of our approach.