基于聚类的非中心化工作流的适应性切片方法

Clustering based adaptable fragmentation method for decentralized workflow

为提升实例密集型的非中心化工作流的执行效率,综合考虑工作流中的活动执行频次,以及活动间的依赖频次和通信量等因素,提出一种基于聚类的非中心化工作流的适应性切片方法。基于活动间的偏序关系,尽可能将频繁通信且通信量大的活动切分在同一个工作流片段中并分配在同一台服务器上执行,可较好适应执行环境。通过模拟实验与已有相关切片方法进行对比分析,分析结果表明,该方法在一定程度上加快了响应时间并提升了吞吐量。

To improve the efficiency of instance-intensive decentralized workflow execution,considering the frequency of activity execution,the dependent frequency and communication time among activities of workflow,etc.,a clustering based adaptable fragmentation method for decentralized workflow was proposed.Based on the partial order relationship among activities,the frequent interactive activities with large communication data were encapsulated in the same fragment and they were distributed in the same server as far as possible,which was more adaptable to execution environment of workflow.A comparative analysis between the proposed method and other relevant fragmentation methods was conducted by simulation experiments.The results show that the proposed method has certain degree of improvement on response time and throughput.