Intra-Subnet Load Balancing in Distributed Workflow Management Systems

Abstract

For enterprise-wide and cross-organizational process-oriented applications, the execution of workflows (WF) may generate a very high load. This load may affect WF servers as well as the underlying communication network. To improve system scalability, several approaches for distributed WF management have been proposed in the literature. They have in common that different partitions of a WF instance graph may be controlled by different WF servers from different subnets. The control over a particular WF instance, therefore, may be transferred from one WF server to another during run-time if this helps to reduce the overall communication load. Thus far, such distributed approaches assume that exactly one WF server resides in each subnet. A single server per subnet, however, may become overloaded. In this paper, we present and verify a novel approach for replicating WF servers in a distributed workflow management system. It enables an arbitrary and changeable distribution of the load to the WF servers of the same subnet, without requiring additional communication.