Discrete Island-Based Cuckoo Search with Highly Disruptive Polynomial Mutation and Opposition-Based Learning Strategy fo
- PDF / 2,299,205 Bytes
- 21 Pages / 595.276 x 790.866 pts Page_size
- 72 Downloads / 160 Views
RESEARCH ARTICLE-COMPUTER ENGINEERING AND COMPUTER SCIENCE
Discrete Island-Based Cuckoo Search with Highly Disruptive Polynomial Mutation and Opposition-Based Learning Strategy for Scheduling of Workflow Applications in Cloud Environments Noor Aldeen Alawad1 · Bilal H. Abed-alguni1 Received: 28 January 2020 / Accepted: 11 November 2020 © King Fahd University of Petroleum & Minerals 2020
Abstract The optimization-based scheduling algorithms used for scheduling workflows in cloud computing environments may easily get trapped in local optima, especially in the beginning of their simulation processes because of some limitations in their exploration methods. Moreover, the performance of some optimization-based scheduling algorithms may severely degrade when dealing with medium- or large-size scheduling problems. The Island-based Cuckoo Search with highly disruptive polynomial mutation (iCSPM) algorithm is a parallel version of the Cuckoo Search (CS) algorithm. The iCSPM algorithm incorporates the island model into CS and uses an exploration function based on the highly disruptive polynomial mutation. It has been empirically proven that iCSPM performs better than popular optimization algorithms (e.g., CS and island-based Genetic algorithm). This paper presents a variation of iCSPM called Discrete iCSPM with opposition-based learning strategy (DiCSPM) for scheduling workflows in cloud computing environments based on two objectives: computation and data transmission costs. DiCSPM includes two new features compared to iCSPM. First, it uses the opposition-based learning approach (OBL) in the initialization step at the level of islands, where each island in the island model contains the opposite population of another island. Second, the smallest position value method is used in the DiCSPM algorithm to determine the correct values of the decision variables in the candidate solutions. The proposed algorithm was experimentally evaluated and compared to well-known scheduling algorithms [Best Resource Selection, Particle Swarm Optimization (PSO) and Grey Wolf Optimizer] using two types of workflows: balanced and imbalanced workflows. The overall experimental and statistical results indicate that DiCSPM provides solutions for the scheduling problem of workflows in cloud computing environment faster than the other compared algorithms. Moreover, DiCSPM was evaluated and compared to state-of-the-art algorithms, namely PSO, binary PSO and discrete binary cat swarm optimization using scientific workflows of different sizes using WorkflowSim. The obtained results suggest that DiCSPM provides the best makespan compared to the other algorithms. Keywords Cuckoo Search · Island model · Highly Disruptive Polynomial Mutation · Opposition-based learning · Cloud computing · Optimization · Workflow · Scheduling · Load balancing
1 Introduction Cloud computing is a relatively new approach that allows the development of distributed, dynamic and fast software. A workflow is a computerized application that is composed of a set of dependent tasks th
Data Loading...
