Enhancement of Cloud Performance using the Clustering Method of Scientific Workflow Task

The scientific community may save money using cloud computing to roll out massive workflow apps. Scheduling workflow jobs with changing resources is a complex problem when running large-scale scientific applications for workflow in a distributed heterogeneous environment. In addition, a scheduling algorithm has to effectively use the resource’s time in a utility-based computing environment like the cloud, which allows for a pay-per-use model of the resources. The cloud is inherently dynamic, yet most scheduling methods ignore this and instead generate static schedules. A public cloud setting, such as Amazon EC2, has access to a plethora of resources and pays for them hourly. Any time used in fractions of an hour is rounded up to the nearest whole hour. The proposed work centres on integrating an accounting model cognizant task clustering technique into the workflow scheduling procedure to satisfy the customer’s budget and deadline. A method for selecting the most appropriate resources at any given point in a process is also presented. Two scientific programs, Montage and Cyber Shake, were run with varying resources and task clustering methods, and the first findings are described.