分布式任务调度在电力市场交易系统中的应用设计

APPLICATION DESIGN OF DISTRIBUTED TASK SCHEDULING IN POWER MARKET TRADING SYSTEM

为应对电力交易规模不断扩大,电力交易系统已开始引入云计算、微服务、大数据等最新IT技术。使用微服务技术会进一步增加系统的分布式架构特性,这种架构的演进会对系统的任务调度功能提出更高的要求。在分析电力交易系统集群规模小、任务绝对量少、策略复杂等特点的基础上,参考互联网行业的解决思路,提出一个基于分布式控制的任务调度解决方案。利用Leader选举的方式,解决互斥任务的调度,并且有效规避脑裂问题。采用易于同Spring整合的触发机制,降低介入的复杂度,解决新型分布式电力交易系统的任务调度功能面临的难题。通过压力测试和实践证明,该方案具有可实施性。

In order to cope with the continuous expansion of the scale of power trading,the electric power trading system has introduced the latest IT technology,such as cloud computing,micro service and big data.With the use of micro service technology,the distributed architecture characteristic of the system will be further increased,and the evolution of this architecture will set higher requirements for the task scheduling function of the system.In this paper,on the basis of the analysis of the small cluster size of the power trading system,the few task absolute quantity and the complex characteristics of the strategy,a task scheduling solution,based on distributed control,was proposed by referring to the solution of the Internet industry.We used Leader election to solve the scheduling of mutual excluded tasks and effectively avoided the split brain problem.The trigger mechanism,which was easy to integrate with Spring,was adopted to reduce the complexity of the intervention and solve the task scheduling problem of the new distributed power transaction system.The scheme is proved to be feasible through pressure testing and practice.