Multi Criteria based Resource Score Heuristic for Cloud Workflow Scheduling

Scheduling scientific workflows modelled by Directed Acyclic Graphs (DAG) is an NP complete problem. Cloud computing provides reliable Quality of Service defined in terms of Service Level Agreements (SLA). To schedule scientific workflows in cloud environment, where resources are shared, it is important to manage the cloud resources efficiently by maximizing utilization. The dynamic nature of cloud resources, due to sharing, heterogeneity, virtualization and workload variations offer a host of challenges in terms of resource availability and performance. This may have a significant impact on task execution times and data transfer times, thus introducing delay in overall execution time called makespan, In this paper, Multi Criteria based Resource Score Heuristic for Cloud Workflow Scheduling is proposed, with an objective of minimizing makespan considering the probability of temporal availability of resources in a cloud computing environment. To choose the best virtual machine with optimal value of maximum resource availability and minimum task execution time, a method of compensatory aggregation of conflicting criteria, is used for scoring each resource. The simulation results demonstrate that the proposed heuristic, can generate schedules with better makespan minimization and is found to be more reliable since resource availability factor is considered while mapping tasks to virtual machines. (C) 2019 The Authors. Published by Elsevier B.V.