ABS: Agent-Based Scheduling for Data-Intensive Workflow in Software-as-a-Service Environments

Efficient allocation and execution of data-intensive scientific workflows to reduce execution time and the data transferring overhead still remains enormous challenge for the adoption of cloud computing to execute data-intensive scientific workflows. To face such a challenge, in this paper, we introduce an agent-based scheduling approach, namely ABS, that automates the deployment of data-intensive workflows in Softwareas-a-Service (SaaS) environments considering two objectives: makespan and data transfer overhead. The algorithm ABS includes two phases: 1) planning phase, task agents in the same workflow form groups through cooperation to reduce data transfer among tasks, 2) scheduling phase, task groups will be assigned to VMs such that reducing makespans for workflows. The experiments conducted using real-world workflow applications demonstrate that our solution outperform the solution produced by the existing algorithm.