A distributed business process fragmentation method based on community discovery

Abstract

For deploying a traditional centralized business process to distributed systems, how to properly partition the process is a key issue. If closely related sub-process fragments are assigned into many different servers, the frequent interactions of the sub-processes will result in large server communication costs. However, most of the current works in business process fragmentation focus on optimizing load balancing to achieve high execution performance, without considering the intrinsic functional aggregation of partitioned sub-processes. This paper proposes a novel distributed business process fragmentation method based on community discovery, which aggregates functional related process partitions together to optimize server communication costs during business process execution. An advanced fuzzy models (AFM) and control-data flow model which are more suitable for community discovery are presented. And the community-based process fragmentation algorithm (CPFA) is developed to get an optimized business process distributed deployment schema, reducing communications between servers and data access communication costs. We evaluate our method with BPI dataset, and demonstrate that it significantly reduces communication costs between servers for process execution efficiency while also maintain good server resource load balancing.