Design of a new task scheduler for computational grid systems using artificial intelligence Algorithms
Subject Areas :Shima Shahabi 1 , ali harounabadi 2 , Seyed Javad Mirabedini 3
1 - Islamic Azad University, Bushehr Branch
2 - computer , central branch , azad university , tehran , iran
3 - Faculty of Islamic Azad University, Tehran
Keywords:
Abstract :
Computational grid systems are a type of large-scale distributed systems, which is most focused on large-scale resource-sharing. The dynamic nature of grid environment and decentralized nature of its resources makes the presence of a capable scheduler application in these systems an absolute necessity. Tasks scheduling is considered a hard problem, so deterministic algorithms cannot effectively solve it. Therefore, many researchers have modified various heuristic algorithms such as genetic algorithm (GA) to solving this problem. Simplicity and parallel nature of GA, which enables it to search the problem space from several different directions, makes GA a good instrument to solve optimization problems. However, this algorithm is more efficient in global searches and has a poor performance in local ones; therefore it should be combined with local search algorithms in order to eliminate this weakness. In this paper, a hybrid scheduling algorithm, which is a combination of GA and variable neighborhood search algorithm, is proposed to solve the problem of scheduling independent tasks on a grid environment. In the proposed algorithm, two factors of flow time and makespan have optimized simultaneously and the cost and number of missed tasks are also considered. The running time of the algorithms proposed algorithm has been improved compared to the 0.18.
