A case study of task priority effects in GA for cluster based DAG scheduling

Directed acyclic graph (DAG) has been widely used in parallel and distributed applications. It is made up by the tasks that are decomposed from the complex application processes. To reduce the intra-tasks dependency time and the total workflow processing time, a promising design named cluster based schedule was put forward. The most popular way for generating the clusters is the genetic algorithm (GA) based method. However, in the GA based clustering processes, different priorities calculation methods lead to totally different results. Aiming at having a deep cognition of the effects of the priorities, this paper comparatively studied the tasks schedule performance with different priorities calculation methods. The experiment results show that the longest path priority can lead to a better schedule performance compared with the other ones.