Energy Efficient and Reliability Aware Workflow Task Scheduling in Cloud Environment

摘要

With the rising demand for cloud services, the high energy consumption of cloud data centers is a significant problem that needs to be handled. The Dynamic Voltage and Frequency Scaling approach has been identified as one of the efficient techniques to conserve energy, particularly while scheduling real-world scientific workflows. Moreover, scientific workflows demand high-availability of the system. The computational systems in the cloud data centers are not failure-free and further frequency scaling impacts negatively on the reliability of the system by increasing the transient fault rate. A trade-off is required between energy conservation and the reliability of the computational machine. In this paper, we propose an energy-efficient and reliability aware workflow task scheduling in a cloud environment (EERS) algorithm, which conserves energy and maximizes the system reliability. The EERS comprises five sub-algorithms. First, we apply a task rank calculation algorithm to preserve the task dependencies. Second, a task clustering algorithm to reduce the communication cost which reduces energy consumption. The third is the sub-target time distribution algorithm to define the sub_makespan for each task. Further, we propose a cluster-VM mapping algorithm that reduces energy and maximizes system reliability and finally, a slack algorithm to reclaim slack associated with the non-critical tasks. The performance of the EERS evaluated on the WorkflowSim simulator using two real-world scientific workloads CyberShake and Montage. The results indicate that it surpasses the related existing approaches.