MODE: A Multi-Objective Strategy for Dynamic Task Scheduling through Elastic Cloud Resources
Authors
Abstract
Cloud computing is introduced as a high-performance computing environment that manages a variety of virtualized resources. One of the major aspects of cloud computing is its dynamic scheduling of great number of task requests that are submitted by users. Cloud data centers in addition to implementing these tasks, should meet the conflicting multiple requirements of different users. Minimizing makespan and deadline violation on a great number of tasks are difficult while costs are reduced. Therefore, in this paper, a multi-objective strategy for dynamic task scheduling through elastic cloud resources (MODE) is proposed, where an algorithm is proposed to construct individual non-dominated sets of new received tasks. These non-dominated sets are sorted in different levels through a new crowding distance of the individuals. In addition, an elastic resource provisioning based on the maximum available VMs’ load is proposed to provide resources in a dynamic manner. The total cost, makespan, and the deadline violations are reduced by 85.84%, 58.03%, and 47.77%, respectively, and the utilization of virtual machines is increased up to 53.2% through this strategy when compared to its counterparts.
