Load Balance with Imperfect Information in Structured Peer-to-Peer Systems

• Main Authors: Ssu-TaChen, 陳思達
• Other Authors: Hung-Chang Hsiao
• Format: Others
• Language: en_US
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/25079580637175719285

碩士 === 國立成功大學 === 資訊工程學系碩博士班 === 98 === With the notion of virtual servers, peers participating in a heterogeneous, struc- tured peer-to-peer (P2P) network may host di?erent numbers of virtual servers, and by migrating virtual servers, the peers can balance their loads proportional to their capacities. The existing and decentralized load balance algorithms designed for the heterogeneous, structured P2P networks either explicitly construct auxiliary networks to manipulate global information or implicitly demand the P2P substrates organized in a hierarchical fashion. Without relying on any auxiliary networks and independent of the geometry of the P2P substrates, we present in this thesis a novel load balanc- ing algorithm that is unique in that each participating peer is based on the partial knowledge of the system to estimate the probability distributions of the capacities of peers and the loads of virtual servers, resulting in imperfect knowledge of the system state. Having the imperfect system state, the peers compute their expected loads and reallocate their loads in parallel. Together with the rigorous performance analysis for our estimation of the system state, we assess our proposal through extensive simula- tions. The simulation results reveal the following: (1) our design is comparable with the centralized solution and outperforms the hierarchical approach in terms of the load imbalance factor, the movement cost of virtual servers, and/or the protocol message overheads; (2) while the existing solutions introduce hotspots to the system due to the manipulation of their load balancing algorithms and thus generate another load imbalance issue, each peer in our proposal experiences the nearly identical workload in performing our load balancing algorithm; and (3) our proposal adapts well to dynamic environments in which peers may come and go freely, and/or the capacities of partici- pating peers and the loads of virtual servers vary over time.