Weighted-adaptive Inertia Strategy for Multi-objective Scheduling in Multi-clouds

Abstract

One of the fundamental problems associated with scheduling workflows on virtual machines in a multi-cloud environment is how to find a near-optimum permutation. The workflow scheduling involves assigning independent computational jobs with conflicting objectives to a set of virtual machines. Most optimization methods for solving non-deterministic polynomial-time hardness (NP-hard) problems deploy multi-objective algorithms. As such, Pareto dominance is one of the most efficient criteria for determining the best solutions within the Pareto front. However, the main drawback of this method is that it requires a reasonably long time to provide an optimum solution. In this paper, a new multi-objective minimum weight algorithm is used to derive the Pareto front. The conflicting objectives considered are reliability, cost, resource utilization, risk probability and makespan. Because multi-objective algorithms select a number of permutations with an optimal trade-off between conflicting objectives, we propose a new decisionmaking approach named the minimum weight optimization (MWO). MWO produces alternative weight to determine the inertia weight by using an adaptive strategy to provide an appropriate alternative for all optimal solutions. This way, consumers' needs and service providers' interests are taken into account. Using standard scientific workflows with conflicting objectives, we compare our proposed multi-objective scheduling algorithm using minimum weigh optimization (MOS-MWO) with multi-objective scheduling algorithm (MOS). Results show that MOS-MWO outperforms MOS in term of QoS satisfaction rate.