EWS: A Pattern Prediction-Based Elastic Workflow Service in the Cloud

Workflows have been widely used to coordinate business activities, software components and services. By making use of cloud resources, workflow systems can be deployed as common services to process a large quantity of workflow execution requests from various users. A challenge for the workflow service is that it should be elastic, i.e., to scale up or scale down to meet the requirements of processing dynamic and changing workloads efficiently while not wasting resources. To address this issue, we implemented an Elastic Workflow Service (EWS). Specifically, we propose a pattern prediction-based strategy in EWS which can dynamically deploy or undeploy components onto cloud resources according to the workload levels. The experimental results show that our configuration strategy has better performance when compared against other strategies.