Load Balancing in strongDynamicstrong Structured P2P strongSystemsstrong.PDFVIP

Load Balancing in Dynamic Structured P2P Systems Brighten Godfrey Karthik Lakshminarayanan Sonesh Surana Richard Karp Ion Stoica {pbg, karthik, sonesh, karp, istoica}@cs.berkeley.edu Abstract— Most P2P systems that provide a DHT abstraction Several solutions have been proposed to address the load distribute objects randomly among “peer nodes” in a way that balancing problem [2], [5], [6]. However, these all assume that results in some nodes having Θ(log N ) times as many objects the system is static and most assume that the IDs of both nodes as the average node. Further imbalance may result due to non- uniform distribution of objects in the identifier space and a and items are uniformly distributed. In this paper, we present a high degree of heterogeneity in object loads and node capacities. solution for a system in which these assumptions do not hold. Additionally, a node’s load may vary greatly over time since the In particular, we consider a system in which system can be expected to experience continuous insertions and deletions of objects, skewed object arrival patterns, and continuous • data items are continuously inserted and deleted, arrival and departure of nodes. • nodes join and depart the system continuously, and In this paper, we propose an algorithm for load balancing in • the distribution of data item IDs and item sizes can be such heterogeneous, dynamic P2P systems. Our simulation results show that in the face of rapid arrivals and departures of objects skewed. of widely varying load, our algorit