Fair-Share Methods for Scheduling Scientific Workflows in Cloud

Efficient execution of scientific workflows in cloud environments, specifically using virtual machines (VMs) necessitates the optimal allocation of resources. However, striking the delicate balance between over-renting and under-renting resources poses a significant challenge, impacting cost and execution time. To address this issue, this paper introduces two fair-share-based scheduling methods that bridge the gap between scheduling scientific workflows in the cloud and traditional fairshare scheduling problems. These methods encompass three crucial steps: partitioning, resource allocation, and scheduling. The workflow is divided into distinct partitions during the partitioning step, with computational requirements calculated for each partition. The resource allocation step distinguishes the two methods, wherein a fair distribution of VMs among partitions is pursued. The proposed methods employ the strategies of max-min and proportional fair share to ensure fair resource allocation. Subsequently, the scheduling step determines the execution schedule based on the allocated resources. Experimental results demonstrate the significant improvements achieved by the proposed methods, enhancing resource utilization and minimizing the elusive concept of makespan. This research paper describes the complexities associated with resource management in cloud-based scientific workflows and provides practical solutions to optimize execution efficiency in these environments.