Balanced Distribution of Load on Grid Resources using Cellular Automata

Load balancing is a technique for equal and fair distribution of workloads on resources and maximizing their performance as well as reducing the overall execution time. However, meeting all of these goals in a single algorithm is not possible due to their inherent conflict, so some of the features must be given priority based on requirements and objectives of the system and the desired algorithm must be designed with their orientation. In this article, a decentralized load balancing algorithm based on Cellular Automata and Fuzzy Logic has been presented which has capabilities needed for fair distribution of resources in Grid level. Each computing node in this algorithm has been modeled as a Cellular Automata's cell and has been provided with the help of Fuzzy Logic in which each node can be an expert system and have a decisive role which is the best choice in a dynamic environment and uncertain data. Each node is mapped of one of the VL, L, VN, H and VH state based on information exchange on certain time periods with its neighboring nodes and based on fuzzy logic tries to decrease the communication overhead and estimate the state of the other nodes in subsequent. The decision to send or receive the workload is made based on each node state. Thus, an appropriate structure for the system can greatly improve the efficiency of the algorithm. Fuzzy control does not search and optimize, just makes decisions based on inputs which are effective internal parameters of the system and are mostly based on incomplete and nonspecific information. Each node based on information exchange at specific time periods with its neighboring nodes, and according to Fuzzy Logic rules is mapped of one of the VL, L, N, H and VH states. To reduce communication overhead, with the help of Fuzzy Logic tries to estimate the state of the other nodes in subsequent periods, and based on the status of each node, makes a decision to send or receive workloads. Thus an appropriate structure for the system can improve the efficiency of the algorithm. In fact, Fuzzy Logic does not search and optimize, just makes decisions based on the input parameters which are often incomplete and imprecise.