Research And Structure Design Over Load Balacing Based On Consistent Hashing With Bounded Loads

With the rapid development of mobile internet technology,clusters of internet servers are facing unprecedented challenge,which makes new suitable implementations of load balancing system especially urgent.Consistent hashing with bounded loads(CHBL)was introduced as the allocation algorithm in the load balancing system of the server cluster to allow the best executive performance,as well as the dynamic adjustment of the cluster.This paper makes following research and optimization based on CHBL.Firstly,Adaptive Consistent Hashing with Virtual Nodes(ACHVN)is proposed in this paper,which optimizes the CHBL in the sector of performance and adaption by applying the utilization rate of computational resources and virtual nodes.The ACHVN could provide better performance in the efficiency of allocation and support for heterogeneous clusters,as well as the capability of dynamic adjustment.Secondly,Genetic Backward Propagation Neural Network with Adaptive Learning Rate is proposed under the scenario of adjusting the weights of computational resources in ACHVN,which resolves the problem that the weights are difficult to scale and analyze.By applying genetic algorithm and adaptive learning rate to BP neural network,the optimized weights of computational resources provided by independent nodes for neural network could reflect the requirement of loads for computational resources better.Finally,this paper presents a structure design for micro-service structure based on ACHVN,which contains a network with three layers consisting of service allocation nodes,load balancing nodes for single service and service nodes,elastically splitting the computational requirement of each load balancing node.Compared with CHBL by experiment,the ACHVN could provide better performance on allocation and possesses the adaption in heterogeneous clusters,whose structure design could perform better load balancing on contrast to conventional Nginx server.