多计算中心作业迁移系统JMS的设计与实现

Design and Implementation of Job Migration System (JMS) in Multiple Computing Centers

摘要: 各个高性能计算中心之间缺少统一的作业管理策略和作业迁移机制。当遇到计算中心例行维护或者用户采用检查点技术进行异地作业灾备容错时，需要将作业迁移到其他计算中心执行。针对该问题，设计实现了多中心作业迁移调度系统JMS。该系统采用了微服务架构的设计模式，支持多计算中心间的作业迁移，为用户提供了多中心作业管理和终端访问的web portal。实际测试表明，JMS具有操作简单，可扩展性强等特点。

Abstract: There is a lack of unified job management strategy and job migration mechanism between high performance computing centers. When there is a routine maintenance in a computing center or the user adopts the checkpoint technology to make fault tolerance, jobs need migrating to other computing centers for execution. Aiming at the problem, this paper proposed a JMS system which adopts the design mode of microservice architecture, supports job migration in multiple computing centers and provides the function of multi-center job management and terminal access for users by web portal. Theoretical analysis and test show that JMS is user-friendly and has good scalability.