Research And Implementation Of A Similarity-based Resource Scheduling Method For Microservice-based Applications In Clouds

With the increasing development and maturity of microservice software architecture,people have clearly realized that whlie microservice architecture brings the advantages of low coupling and high availability to software system,it also increases the system cost of communication and cooperation between services,and also increases the cost of software system deployment and operation.Therefore,how to effectively use system resources and ensure system service quality has become a common concern in the field.At present,most of the solutions are based on cost for quality.According to the business scenarios and domain experience,the system is deployed according to the scheme that can meet the peak demand.However,user requests for system in the cloud environment are continuous,changing and unpredictable,which determines the current single deployment mode,and cannot effectively solve the above problems.Therefore,the distributed deployment state of microservice system also needs to continuously change with the change of system load(user requests).How to ensure high service quality,fast service response speed and less system resources is an urgent problem to be solved.To solve the above problems,this paper proposes a resource scheduling algorithm for cloudbased microservice systems based on similarity.First,the performance and stability of the entire microservice system is modeled.Then,based on the performance model,we set the objective function for the scheduling algorithm,and compared it with three other classical heuristic algorithms to draw the conclusion.This paper is dedicated to being able to target microservice systems in the cloud environment,with the help of lightweight container technology,and dynamically change the system’s deployment state based on changing system loads,thereby improving the service performance of the entire microservice system and reducing resource consumption.The specific research content includes the following aspects:(1)Based on the micro-service system in the cloud environment,modeling the performance and stability of the whole system.Specifically,the system is modeled from three aspects: the resource cost of the entire system in the cloud,the instability brought to the system when the system deployment state changes,and the calling cost when microservices call each other,covering the entire system’s resources cost,performance,stability and more.(2)Tea Store,a current relatively new and standard microservices benchmark system,was used it as the test environment to deploy the system on the cloud environment,and studying the association between various microservices in Tea Store.We conducted data statistics on the relationship between microservices and the occupied resource footprint of Tea Store in order to apply the three target models in(1)to the Tea Store system.(3)Designing and implementing the based on similarity dynamic resource scheduling algorithm(Similarity-based Dynamic Resource Sheduling Algorithm for Microservicebased Web Systems,Sim-DRS).Aiming at the dynamic and real-time nature of microservice resource scheduling on the cloud,non-dominant sorting genetic algorithm II algorithm and k-means clustering algorithm were introduced in the population initialization process to solve the problems caused by the random initialization method used by ordinary evolutionary algorithms and improve the performance of sim-drs algorithm.This article uses the open source Tea Store system to verify the based on similarity dynamic resource scheduling algorithm(Sim-DRS).Under the strict time constraint,we compare the Sim-DRS algorithm with the other heuristic algorithm,ACO、PSO and NSGA II,finally come to the conclusion: in the case of more stricter time constraints,the effect of the SimDRS algorithm is better than that of the three comparison algorithms,which achieves our optimization purpose.