Parallel Particle Swarm Algorithm-Based Optimization Model For Microservice Container Scheduling

With the rapid growth of the Internet industry,the deployment of containers in cloud architectures has become increasingly common.This is due to their ease of deployment,minimal overhead,and high portability.When applications are containerized,multiple containers are deployed on each physical node,either on the same node or different nodes,forming a consumption relationship upon receiving user requests.However,issues such as load balancing,network latency,and user requests arise within this consumption relationship,making the deployment of containers to the appropriate locations a critical concern.The optimization of multiple objectives simultaneously presents a classic multi-objective optimization problem.Some researchers have improved upon the default algorithms of scheduling tools to achieve more desirable results,while others have proposed their own multi-objective optimization models and algorithms tailored to the current context.However,the microservice container scheduling model still suffers from incomplete considerations,low optimization efficiency,and high memory consumption.To address these challenges,this paper introduces a new objective optimization model,MMO-MS,which incorporates realistic usage factors such as network latency,request failure rate,and load balancing requirements in different scenarios.Additionally,it enhances the model with a parallel particle swarm algorithm specifically designed for microservice scheduling,combined with the probabilistic mutation feature of simulated annealing algorithms.Simulation experiments are conducted to compare the results with other classical algorithms,and the experimental findings demonstrate that the improved algorithm in this paper yields superior optimization results for the MMO-MS model.(1)MMO-MS model: Building upon the analysis of models proposed by previous researchers,this paper focuses on modeling two aspects: usage scenarios and the relative location differences of containers.The model addresses local load balancing by considering multiple physical resources,allowing for flexible adjustments when optimizing microservices in various application scenarios.In terms of network transmission overhead,the paper emphasizes latency,which is the most critical factor impacting user experience during real usage.Regarding service reliability,the model considers not only the relative location and number of request failures but also incorporates other factors.Finally,these three optimization objectives are integrated into the new MMO-MS model.(2)SAPPSO-MCS algorithm: Once the corresponding optimization objectives are established,finding the best container scheduling solution in a goal-oriented manner becomes crucial.While MOAPPS-MCS successfully tackles this problem,it suffers from excessive data exchange between populations during parallel exchange.To address this issue,this paper enhances the MOAPPS-MCS algorithm by incorporating the probabilistic features of simulated annealing.The results,when combined with the multi-objective optimization model,demonstrate significantly improved optimization efficiency compared to previous approaches.