Abstract
This paper studies the load balancing problem in cloud computing, which was the topic of the 2012 ROADEF/EURO challenge jointly organized by the French Operational Research and Decision Support Society (ROADEF) and the European Operational Research Society (EUOR). The aim of the challenge was to concentrate research efforts to solve the most important optimization problems in applied fields. The subject of this challenge was a key problem that was encountered by Google, related to the manner of scheduling a series of computing tasks in a large-scale cluster system. The problem is a bottleneck problem in cloud computing, and has shown to be NP-hard. Research on the problem can help cluster systems perform computing tasks more efficiently, improve the usage effectiveness of servers, and reduce system congestion. The objective of the problem is to reassign machines to processes in order to meet as many soft constraints as possible while satisfying all hard constraints. We propose a local search-based metaheuristic algorithm to solve this problem. The algorithm consists of three neighborhood structures, a dynamic perturbation strategy, an efficient partition mechanism for search space, and a cache technology to speed up the search. We tested the algorithm on a total of 20 instances of A and B sets used in the challenge. Computational results showed that scores obtained by the algorithm were 16% and 4.23% for the instances of the A and B sets, respectively, demonstrating the effectiveness of the proposed algorithm. Furthermore, we compared and analyzed the key components of our algorithm with those of the exact method and the lower bound algorithms.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call