This article studies the application of multiobjective evolutionary algorithms for solving the energy-aware scheduling problem of workflows in a distributed system that is composed by a federation of datacenters. Nowadays, energy efficiency is a major concern when using large distributed computing systems, including novel grid and cloud computing facilities. Researchers and system planners are looking for accurate methods to be used for planning the execution of large workloads that consume large amounts of resources, having a direct implications for the energy consumption of the system and its operational costs. In the approach proposed in this article, we study the application of multiobjective evolutionary algorithms combined with lowlevel backfilling heuristics for finding efficient mappings of workflows into resources in order to maximize several metrics related to the quality of service, while reducing the energy required for computation. The experimental evaluation is performed considering both medium and large workloads that model realistic highperformance computing applications and modern distributed computing infrastructures. The experimental results demonstrate that the proposed multiobjective evolutionary approaches compute accurate schedules, significantly outperforming both traditional round-robin/load-balancing schedulers and a set of combined list scheduling heuristics (accounting for both problem objectives) previously applied to the problem.
Abstract. This article presents sequential and parallel metaheuristics to solve the virtual machines subletting problem in cloud systems, which deals with allocating virtual machine requests into prebooked resources from a cloud broker, maximizing the broker profit. Three metaheuristic are studied: Simulated Annealing, Genetic Algorithm, and hybrid Evolutionary Algorithm. The experimental evaluation over instances accounting for workloads and scenarios using real data from cloud providers, indicates that the parallel hybrid Evolutionary Algorithm is the best method to solve the problem, computing solutions with up to 368.9% profit improvement over greedy heuristics results while accounting for accurate makespan and flowtime values.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.