| |
Click here for full text:
Load Balancing in Distributed Workflow Management System
Jin, Li-jie; Casati, Fabio; Sayal, Mehmet; Shan, Ming- Chien
HPL-2001-287
Keyword(s): load balancing; workflow; load index; business process
Abstract: Workflow Management Systems (WFMS) play a very important role in constructing today's e-commerce environment through automating intra-enterprise business processes and inter-enterprise services. To handle the rapidly changing business environment and global competition, WFMSs should have flexibility and scalability to meet the business requirement and to quickly introduce new and efficient business services. Achieving load balancing is essential to ensure scalability in a distributed WFMS. In this paper we discuss load-balancing technology for distributed WFMSs. First, we introduce a workflow load index to measure load level of workflow engines. Then we present a WFMS cluster architecture with a load balancing subsystem. We compare the performance of round robin versus load-aware scheduling under the same load pattern. The experimental results show that the load index that we define in this paper is a good indicator of the load level in a distributed WMFS. The results also suggest that the load-aware scheduling algorithm can distribute workload fairly on heterogeneous WFMSs; instead, the round robin scheduling can only guarantee load balance in uniform WFMS with uniform workload and resource capabilities. Notes: Copyright ACM. Published in the ACM Symposium on Applied Computing (SAC 2001) 11-14 March 2001, Las Vegas, NV
9 Pages
Back to Index
|
printable version
