摘要
针对结构化P2P系统提出了基于事件驱动的拓扑匹配算法。在该算法中,节点加入系统时依据系统全局信息找到与该节点最近的节点作为邻居节点;当有节点离开或者加入系统从而导致系统覆盖网络拓扑结构发生变化时,只有受影响的节点做出适当的调整。事件驱动算法不是只利用系统全局信息或者只利用系统局部信息,而是二者的结合,从而减少了算法带来的开销。在CAN系统上的仿真试验表明事件驱动算法可以使系统的平均逻辑链路延迟/平均物理链路延迟降低65%左右,而用SAT-match与界标簇算法结合的方法只能降低50%左右。同时,算法带来的开销也是非常小的。
An event-driven based method was proposed to solve the topology mismatching problem of structured P2P systems. In this method, when joining a structured P2P system, based on the global information of the system, a peer is guided to find a physically close neighbor to connect with. If the overlay topology changes as peers join or leave the system, only affected peers are adjusted. The method neither solely relies on global information nor solely relies on local information of the system, but a good cooperation. This character reduces the overhead this method caused. Intensive simulation experiments on CAN overlays show that the event-driven algorithm can achieve average logical/physical link latency reduction rate by up to 65%, outperforming the combination of SAT-Matching and landmark binning which can achieve the reduction rate by up to 50%. Meanwhile, the overhead caused by this method is very small.
出处
《系统仿真学报》
EI
CAS
CSCD
北大核心
2006年第5期1181-1185,共5页
Journal of System Simulation
基金
国家自然科学基金资助(60403031
90604015)
国家863项目(2005AA121560)
关键词
P2P
拓扑匹配
事件驱动
界标簇
P2P
topology matching
event-driven
landmark binning
