Load Balancing Algorithm of GPU Based on Genetic Algorithm

Zhang Xiang Yang (PetroChina Research Institute of Petroleum Exploration &Development-Northwest)

Indonesian Journal of Electrical Engineering and Computer Science
Vol 12, No 6: June 2014

As the development of GPU/CPU parallel computing in recent years, load balance of GPU server has been more and more important, so we promote a genetic algorithm-based load balancing algorithm for GPU RTM. The algorithm takes the server status and job assignment into account, and design a coding mechanism and genetic manipulation, as well as fitness function. The experiments show that,the algorithm can reach a better effect of efficiency and load-balancing. It can hidden data transmission in the parallel computing, and during server downtime, it can prevent the idle of other computing resources. DOI : http://dx.doi.org/10.11591/telkomnika.v12i6.4531