A t-level driven search for estimation of distribution algorithm in solving task graph allocation to multiprocessors

Abstract

The development of cloud computing drives the research on parallel processing. One of the important problems in parallel processing is to minimize the makespan of the tasks with precedence constraints on multiprocessors scheduling. In this paper, the property of the t-level (top-level) is analyzed, and a t-level (top level) driven search is proposed to enhance the exploitation ability of the efficient estimation of distributed algorithm (eEDA), which was developed for solving the precedence constrained scheduling problem. Numerical tests and comparisons are carried out. The results demonstrate that the t-level driven search is able to improve the optimization capacity of the eEDA under heterogeneous multiprocessor situation. Moreover, it is also shown that the eEDA with the t-level driven search on homogeneous computing systems is effective.