Research On Load Balancing And Cache Management Strategy For Proxy Cache Cluster

The geometric growth of Internet access has led to the increased web server load and user access delay.Proxy cache cluster(PCC)is a group of proxy servers with caching capabilities,which are deployed on the edge of backbone network and connected with clients through a high-speed LAN.It can reduce the load of web server and user access delay to a certain extent when PCC is used as middleware between clients and web server.At the same time,it introduces new problems such as load balancing and cache management,which will result in PCC performance degradation if they can not be solved.Therefore,a load balancing strategy based on virtual nodes and a cache management strategy based on generation and comprehensive resident value are proposed to solve these problems.1.Virtual nodes are introduced into consistent hashing with bounded loads strategy to improve its load balancing and make it compatible with homogeneous and heterogeneous PCC.By combining with different performance parameters,performance ratio can be formulated with independent information data fluctuation weighting method,and virtual nodes attached to every cache node can be generated by a random function.The greatest common divisor m of number of virtual nodes attached to every cache node can be calculated and the hash ring is cut into m arcs,and the virtual nodes are mapped proportionally to arcs of the hash ring by MD5 function.Simulation results show that the strategy can effectively improve the load balancing and cache hit ratio in homogeneous and heterogeneous PCC without cache space limitation,and reduce backbone traffic consumption and PCC response time.2.PCC performance is also closely related to cache space capacity.A cache management strategy based on generation and comprehensive resident value is proposed to improve cache and byte hit ratio in limited cache space.In the strategy,cache space is divided into intervals,and independent information data fluctuation weighting method is used to calculate the comprehensive residence value of each cache object.When the capacity of cache interval is insufficient,the generation cache management mechanism of the corresponding interval will be triggered.Cache objects with lower comprehensive resident value in this interval will be removed or “downgraded”,and the remaining will be “upgraded”.Comprehensive tests show that the strategy can guarantee the performance indexes of PCC to be still effectively imporoved in the case of limited cache space.Starting with PCC load balancing and cache management issues,a load balancing strategy named consistent hashing with bounded loads and virtual nodes and a cache management strategy based on generation and comprehensive resident value are proposed separately.A test platform which is close to actual engineering application is built for comprehensive performance test.Test results show that the proposed two strategies can effectively improve PCC load balancing,cache and byte hit ratio,backbone traffic comsumption and PCC response time.