Load Balancing Mechanism Based On Microservice

In the distributed cloud system based on Microservice framework,load balancing is its core component.The operation efficiency of cloud system is directly determined by load balancing strategy and ability.The famous CPA theory puts forward that in the distributed system,consistency,availability and partition tolerance correspond to the three indexes of the distributed system.Due to the complexity of the distributed system,it can not be realized at the same time,in which availability is the key problem discussed in this thesis.Generally,there are two ways to optimize availability,that is,to improve the hardware capability of the server,but it is often difficult to achieve considering the cost,or to balance the load of each node in the system through efficient load balancing strategy.In distributed systems,load balancing,as the core part,has been widely used in various cloud systems.The typical case of distributed system is micro service architecture.It divides the system into multiple groups of small micro services according to different business functions.The coordination and cooperation between different services greatly improves the system performance,but it will also lead to the problem of service load imbalance,that is,the amount of partial service requests is too high and the amount of partial service requests is too low.Aiming at the load balancing strategy,this thesis completes the following work:(1)This thesis studies the load balancing mechanism under the current microservice architecture,and analyzes the advantages and disadvantages of the traditional load balancing mechanism.Most of the popular microservice systems use static load balancing algorithms,which can not carry out dynamic operation according to the operation status of service instances,resulting in load tilt when the concurrency surges.The traditional dynamic load balancing algorithm does not have the ability of service governance,which is easy to cause the problems of load tilt and service overhead expansion.(2)Research and analyze the advantages and disadvantages of the current popular load balancing strategy,and make targeted improvements.In this thesis,a dynamic load balancing algorithm based on load bounded consistent hash is proposed,and the concepts of system resource utilization and virtual node are introduced to optimize it.Experiments show that it has better allocation efficiency and dynamic scalability than the traditional algorithm,and ensures the high availability of the distributed system.(3)The improved dynamic consistency hash algorithm is applied to the actual microservice architecture scenario.In the system,the load balancer used to encapsulate the algorithm is analyzed and designed according to the actual application scenario,so as to realize the dynamic scalability and high availability of the distributed system.At the same time,this thesis will also adjust the load balancing algorithm,conduct secondary modeling combined with the micro service architecture system,and optimize the load balancing strategy structurally.(4)The experimental scheme is analyzed and designed,and the load balancing algorithm is compared in many aspects.Compared with the traditional load balancing algorithm,the improved dynamic consistent hash algorithm has better allocation efficiency and reduces the load imbalance rate by about 25%.In addition,experiments are carried out on the redesigned load balancing system,which also has better performance than the traditional load balancing server.