面向交叉微服务链的任务调度优化

Optimization of task scheduling oriented to cross microservice chains

微服务架构将应用程序划分为一组松散耦合的细粒度服务,各个微服务独立部署更新。同时,这些微服务相互协作形成了多条存在交叉的微服务链,服务的交叉点成为资源竞争的关键位置。因此,合理地为服务分配资源,能有效地解决服务链交叉带来的资源竞争问题,从而提高任务调度过程中的资源利用率,降低了任务响应时间。然而现有的研究中往往忽视或简化了服务链交叉访问微服务时产生的冲突问题,导致系统调度效果差。为此,针对微服务链交叉产生的资源竞争问题,以系统资源利用率及处理请求的全局响应时间为衡量指标,将微服务架构中服务的资源消耗情况和任务执行的时间进行了形式化表征。同时,结合蚁群算法并行计算与模拟退火算法局部扰动的优势,提出了一种面向交叉微服务链的任务调度算法。通过实验证明,与先来先服务算法和传统蚁群算法相比,文中的算法能够在复杂微服务链环境下有效提高资源利用率,并降低任务的全局响应时间。

The microservice architecture arranges an application as a set of loosely-coupled fine-grained services,with each microservice independently deployed and updated.The cooperation of services leads to multiple intersecting microservice chains.And the intersection of microservices becomes a key position for resource competition.Therefore,rational allocation of microservices can improve resource utilization,reduce the task response time and solve the problem of resource competition caused by the intersection of microservice chains.However,existing research often ignores or simplifies the conflict problem caused by the intersection of microservice chains,resulting in poor system scheduling.Therefore,aiming at the above problem,this paper takes the resource utilization and the global response time as the measurement indicators to formally characterize the resource consumption of services and the task execution time in the microservice architecture.Combined with the advantages of parallel computing of the ant colony algorithm and local perturbations of simulated annealing algorithm,this paper proposes a chain-oriented task scheduling algorithm(COTSA).Experimental results show that compared with first come first service(FCFS)and ant colony optimization(ACO),the COTSA can effectively improve resource utilization and reduce the overall response time in the complex microservice environment.