A comparative study of two different optimization methods applied to the static job scheduling problem in grid computing
Régis Donald Hontinfinde, Ariel Kamoyedji, Marcos Thyrbus Vitouley, Roland Smako Honfo · 2023
In the telecommunications sector where real-time data processing and low latency are essential, grid computing promotes faster response times and reduced network congestion. Grid Computing has since become a very powerful and indispensable tool for telecommunications operators because it allows companies to optimize their operations, improve the performance of their network and provide better services to their customers. One of the distributed computing systems used in telecommunications is the User-PC (UPC) system. The User-PC Computing (UPC) system provides a cost-effective distributed computing platform by utilizing the unused resources of personal computers (PCs). Effective job allocation methods are essential for optimizing system performance. In this study, we introduce a comparative analysis of the static job scheduling problem in the UPC system, focusing on the performance comparison between our proposed hybrid approach, which combines a greedy algorithm and a algorithm genetics, and the approach based on Randomized Multi-Start Local Search. Our hybrid algorithm is composed of two steps. In the first step, a Greedy Algorithm is used to create an initial population of candidate solutions by assigning jobs to available workers based on CPU core utilization. In the second step, a Genetic Algorithm is used to optimize job-worker assignments through selection, crossover and mutation operations. To evaluate the performance of our approach, extensive experiments were conducted with up to 27 jobs and 6 worker PCs. The results show that our hybrid approach achieved a shorter makespan compared to the approach based on Randomized Multi-Start Local Search. Specifically, our algorithm outperformed the multi-start random local search-based method by more than 31% in terms of makespan reduction. The experimental results highlight the effectiveness of our hybrid approach to improve job scheduling performance in the UPC system.