A Game Theoretical Load Balancing And Optimization Approach For Distributed System

5G aims to achieve substantial improvement for the performance of wireless network in terms of e.g.higher throughput and lower latency.The rapid development of 5G has also affected the architectural upgrade of distributed systems.Edge computing is the next-generation distributed system that brings the computing and storage capability closer to the terminal users.Smart base stations are often attached with edge datacenters consisting of many edge servers equipped with different computing and storage capabilities,which is the biggest difference between edge distributed systems and traditional distributed systems.Edge servers are used to execute offloaded tasks from edge equipment such as Internet of Things and Virtual Reality.It is important to have an efficient offloading algorithm that can guarantee specific service-level objectives(SLOs)by assigning tasks to appropriate edge servers which are heteromerous in computing and storage capability.Traditional offloading schemes such as static and learning-based algorithms either have limited performance or result in high overhead for task assignment to servers.We propose a lightweight algorithm for load balance based on public good model.The game model focuses on the assigning load weights for the edge distributed systems in the proposed algorithm.This paper makes the following contributions:1.We model the system,measure the computing and storage state of each edge server through historical server processing speed,queue length,and memory load,and formulate the service blocking potential by virtue of the Erlang C formula.2.We design a game model for load balancing optimization based on the public goods model,and formally propose the optimal solution of the game model;We propose an optimized algorithm based on the Smooth Weighted Round-Robin(SWRR)algorithm to determine the proportion of requests that each server should receive.We analyze the performance advantages of this algorithm.3.We design a load balancing system prototype which supports multiple load balancing algorithms,including the proposed load balancing algorithm based on public good game.We compare the proposed algorithm with the random algorithm,the Round-Rabin algorithm,the power of 2 choices algorithm and deep learning based algorithm.The extensive experimental results show that the proposed game algorithm has significant advantages in terms of task processing speed,task response delay,stability,and energy consumption,while significantly reducing the CPU usage and memory usage by up to3 times.