Feedback guided load balancing in a distributed memory environment

Feedback guided dynamic load balancing is the process of equilibrating the load of all the processors of a system during execution time based on information gathered from them. It provides better execution times by better utilizing the systems resources. In this work we have suggested a new dynamic load balancing model, named Wave Propagation Model, that is based on the diffusion balancing method. The suggested model was implemented in a C language library, the DLBLib, which offers to the programmer the ability to add dynamic load balancing to his applications with minimum effort. Simulations and benchmarks executed using the DLBLib showed that the development of this library can be beneficial for parallel applications as it achieves good performance with minimum cost. It will allow programmers to easily load balance their applications whilst not adding complexity to their codes.