MCPF: Fault-Tolerant Scheduling of Scientific Workflow on Cloud Computing

Abstract Scientific workflow applications entail extensive amounts of tasks and data-sets necessitating systematic processing. Cloud platform is utilized for executing these applications which provide access to extensive amounts of scalable and on demand resources. Running scientific workflow applications on cloud computing experiences a huge amount of failure, i.e., hardware failures, software failures, network failures, etc., due to the large scale heterogeneity and distributed nature. That affects overall workflow execution time, monitory cost, and resource utilization. Numerous fault-tolerance methods are used to resolve and handle failures in cloud computing environment. In this paper, we used the MCPF (Multiple Critical Partitions with Failure) technique. The proposed technique has two phases. In the first phase, the rank of all tasks is calculated by summing the ranks, i.e., downward and upward rank. And then, in the second phase tasks are scheduled based on their ranking on the VMs, which has a lower failure rate. We evaluated the performance of our proposed technique under different conditions using parameters, i.e., makespan and cost. We have compared the results of MCPF with well-known existing HEFT, and RDEARP algorithms. Simulation results obtained through experiments and their comparison with existing techniques lead us to the conclusion that our proposed technique yields better results than existing algorithms.