A Bio-inspired Adaptive Job Scheduling Mechanism on a Computational Grid

A computational grid is a highly dynamic and distributed environment. Unlike tightly-coupled parallel computing environment, high performance computing on the grid is complicated by the heterogeneous computational performances of each node, possible node unavailability, unpredictable node behavior, and unreliable network connectivity. Compared to a static scheduling, an adaptive scheduling mechanism is more favorable and attractive in a grid-computing environment, because it can adjust the scheduling policy according to its dynamically changing computational environment. In this paper, we present a job scheduling mechanism that enable the adaptation of naturally parallel and compute-intensive jobs to clustered computational farms with heterogeneous performance. The kernel of this scheduling technique is a swarm intelligent algorithm, which is inspired from the ants' behavior in a social insect colony. We applied the bio-inspired adaptive mechanism in a simulated computational grid and compared it with static scheduling algorithms. Our results showed good performance, adaptability, and robustness in a dynamic computational grid with respect to its competitors.