An adaptive fault detector strategy for scientific workflow scheduling based on improved differential evolution algorithm in cloud

Abstract

With the increasing popularity and acceptance of cloud computing, it is being applied in services like executing large-scale applications, where cloud environment is selected by the scientific associations to easily execute the computation intensive workflows. However, cloud computing can have higher failure rates due to the larger number of servers and components filled with the intensive workloads. These failures may lead to the unavailability of virtual machines (VMs) for computation. Hence, this issue of fault occurrences can be tolerated by adopting an effective and efficient fault tolerant strategy. The goal of our research in this paper is to develop an adaptive fault detector strategy based on Improved Differential Evolution (IDE) algorithm in cloud computing that can minimize the energy consumption, the makespan, the total cost and, at the same time, tolerate up faults when scheduling scientific workflows. This proposed work applies an adaptive network-based fuzzy inference system (ANFIS) prediction model to proactively control resource load fluctuation that increases the failure prediction accuracy before fault/failure occurrence. In addition, it applies a reactive fault tolerance technique for when a processor fails and the scheduler must allocate a new VM to execute the workflow tasks. The experimental results show that compared with existing techniques, the proposed approach significantly improves the overall scheduling performance, achieves a higher degree of fault tolerance with high HyperVolume (HV) compared with the ICFWS, IDE, and ACO algorithms, minimizes the makespan, the energy consumption and task fault ratio, and reduces the total cost.