Communication partner identification in distributed job workflow execution over the grid

Mobile agent based distributed job workflow execution is a promising paradigm for data intensive collaborative scientific computations over the grid. In this paper, a mobile code collaboration framework (MCCF) for distributed job workflow execution over the grid is described and an algorithm to identify partners for agent communication in MCCF is presented. Previous work of mobile agent communication focuses mainly on agent location and communication. Little work has been done on communication partner identification. In MCCF, a novel subjob grouping algorithm for preprocessing the job workflow´s static specification is developed. The obtained information is then used during runtime to identify partners for agent communication. The mobile agent dynamic location and communication based on this approach limits the number of agents required for communication during the dynamic job workflow execution. The algorithm is evaluated through a comparison study using simulated job workflows executed on a prototype implementation of the MCCF. The results show that the algorithm is scalable and efficient.