A Prediction based Proactive Resource Provisioning Strategy for Multi-Objective Workflow Scheduling in Cloud Computing

In order to manage the workload proactively on the Cloud system during application execution, workload should be predicted through a proper approach and count of resources handled through an auto-scaling system controller. On the other hand, the workflow scheduling requires proper mapping of cloud resources to workflow tasks to efficiently utilize resources and meet different user’s quality of service requirements. Workflow scheduling is NP-complete problem and multi-objective evolutionary algorithms have shown their merit for solving such problem. Most of the works in the literature focused either on dynamic resource provisioning or scheduling algorithms for executing workloads. Based on this deficiency, in this paper, a prediction‑based proactive resource provisioning strategy based on learning vector quantization (LVQ) artificial neural network was proposed to predict the workloads in future and a fuzzy system controller was proposed to compute the proper number of resources to be allocated to the Cloud system. In addition, the multi-objective linear programming scheduling algorithm was proposed to execute workloads effectively on available resources. An evaluation with three kinds of real scientific workflows was performed. The experimental results showed that the proposed approach efficiently reduced execution average cost, and response time along with higher resource utilization in comparison with its counterparts.