Grid computing systems are new and important fields, and they differ from common distributed systems due to resource subscription in large scales, high efficiency and creative applied program. Grid computing system is a wide distributed system. The main and special problem of grid concept is coordinating resource subscription and solving the problem dynamically and multi-distribution programs. Grid computing system is a distributed computing system, and in this system, communication and processors have considerable effects on reliability of grid computing. In this paper, the features of grid computing systems have been described, and some algorithms have been presented to analyze reliability of system and grid program. In this study, an intelligence algorithm has been used to create load balance in grid. Since grid environment is dynamic, firefly algorithm has been used to create load balance in sites. The results of algorithm simulation have been presented by Matlab software, and it is compared with learner particles swarm algorithm, simulated annealing and random algorithm.