Research Of Load Balancing In High Concurrent Network Communication Based On RPC

With the rapid development of Internet technology,tens of thousands or even hundreds of millions of users get access to large web applications day after day.Furthermore,the number of page viewing is increasing exponentially and is much greater especially during the period of holidays.High concurrency means that multiple users request one service at the same time per unit time,resource utilization rate of servers will increase and their processing capacity will descend gradually and even steeply.Faced with a considerable amount of concurrent requests,to maintain the original real-time,smooth,stable user experience,good network communication framework is needed to handle this problem.By analyzing the characters of high concurrency,the response time of request and dynamic monitoring of hardware resources,this paper proposes a strategy of dynamic load balancing based on RPC and designs a mathematical model correspondingly.Then realizing multiple servers' cluster system to carry out concurrent testing experiments,through the comparison with commonly used RPC framework,this strategy proves its feasibility and effectiveness.In addition,an evaluation model for high concurrency is proposed and the results are analyzed.Finally,in order to better satisfy the demand of high concurrency,adjusting variable parameters of the model and retrying several experiments,fitting function and obtaining the optimum solution to achieve model optimization.The main work of this paper is as follows:(1)Studying commonly-used load balancing and RPC frameworks,then understanding their general principles.Specifically,the advantages of Dubbo framework and its load balancing.Experiments are carried out on four kinds of Dubbo's load balancing strategies and data is aggregated and compared.The next step is to study the round robin load balancing which is more suitable for high concurrency.According to the characteristics of RPC,this paper puts forward the response time model of RPC,combining with the fastest response algorithm and the monitoring of hardware resource,catching up with dynamic weight value updating strategy and its corresponding mathematical model.(2)According to the dynamic weight value updating strategy proposed in this paper,putting forward the strategy of dynamic weighted round robin load balancing based on RPC,designing and dividing the core modules,which includes dynamic registration module,data cache module,core monitoring and processing module as well as load balancing module,then describing the relationship between these modules.(3)To achieve these modules mentioned above,building a three-service-provider cluster.By comparing with the RPC original framework,the feasibility of this strategy was verified.It can reduce the average response time and what is more,under conditions of high concurrency,the efficiency of the improved RPC framework was better than that of the usual RPC framework.Also,it can fully utilize the better service provider.(4)Finally,an evaluation model for load balancing efficiency was proposed to optimize the model parameters.Through the adjustment of variable parameters,obtaining the comparison of load balancing performance under different conditions,at the end of this paper,a quadratic fitting function for this study was fitted by the experimental results,and the fitting function of this experiment was obtained by solving the equation and the idealized parameter values are calculated and put forward.