Task Scheduling by Neural Network with Mean Field Annealing Improvement in Grid Computing

Abstract

Task scheduling is a key concern in developing grid computation application. Desirable goals for grid task scheduling algorithms would shorten average delay and maximize system utilization and fulfil user constraints. In this work, an agent-based grid management infrastructure is coupled with Hopfield neural network scheduling algorithm. An agent in a grid utilizes a neural network algorithm to manage and schedule tasks. Hopfield neural network is good at finding optimal solution with multi-constraints and can be fast convergent to the result. The simulation results show that the scheduling algorithm works effectively. Efficient and valid solutions for grid task scheduling can be obtained using the scheme. Hopfield neural network is good at finding optimal solution with multi-constraints and can be fast convergent to the result. However, it is often trapped to a local minimum. Mean field annealing algorithm has an advantage in finding the optimal solution escaping from the local minimum