Blacklist muti-objective genetic algorithm for energy saving in heterogeneous environments

Abstract

Reducing energy consumption in large-scale computing facilities has become a major concern in recent years. Most of the techniques have focused on determining the computing requirements based on load predictions and thus turning unnecessary nodes on and off. Nevertheless, once the available resources have been configured, new opportunities arise for reducing energy consumption by providing optimal matching of parallel applications to the available computing nodes. Current research in scheduling has concentrated on not only optimizing the energy consumed by the processors but also optimizing the makespan, i.e., job completion time. The large number of heterogeneous computing nodes and variability of application-tasks are factors that make the scheduling an NP-Hard problem. Our aim in this paper is a multi-objective genetic algorithm based on a weighted blacklist able to generate scheduling decisions that globally optimizes the energy consumption and the makespan.

Hybrid Metaheuristic Based Offline Parallel Job Scheduling in Heterogeneous Computing Systems

“…During PSO iteration, a fuzzy genetic crossover operator was applied over PSO particles with the objective to minimize makespan and energy consumption simultaneously. MOPSO-FGA failed to produce best makespan and energy consumption when compared with MOGA [18] but it produced shorter algorithm run time than MOGA. Authors in [2015-Switlaski] introduced the Generalized External Optimization (GEO) based scheduling policy for batch of parallel jobs consisting of independent tasks in computational grids to minimize the makespan.…”

“…GEO based policy produced better results over GA policy in all the test cases of three sets of jobs. Authors in [18] proposed Multi-objective Genetic Algorithm (MOGA) based solution using NSGA-II for scheduling of batch of collaborative parallel jobs in federated cluster computing structures to minimize makespan and energy consumption simultaneously. MOGA was used to provide solution for job ordering and processor allocation to jobs.…”

“…Most of the research works listed in this section focused on scheduling either independent jobs or parallel jobs consisting of independent tasks in HCS platforms without considering bandwidth penalties. We found very few research works [15][16], [18][19][20] which provide offline solution for scheduling collaborative parallel jobs by considering computation heterogeneity and inter-cluster communications. Such parallel job's tasks have computation and communication phases and tasks can collaborative with each other during communication phase.…”

“…We follow the execution time model of [38] which is also adapted by many related research works [16] [18][19], to model the execution time slowdown due to heterogeneity of computational resources and communication contention present on shared communication link in multi-cluster HCS platform. For a parallel job j, Cumulative slowdown in the presence of processing slowdown ( ) and communication slowdown ( ), is calculated using (1).…”

“…Since job scheduling problem consists of two sub-problems; Job execution order and resource allocation, therefore each individual S, also known as a candidate solution to the scheduling problem, is composed of job order (O) vector and resource availability (A) matrix components [18]. Let each index of a job order continuous solution represent a unique job from set of n jobs in the multi-cluster system.…”