Analysis and review of effectiveness of metaheuristics in task scheduling process with delineating machine learning as suitable alternative

For any computing environment, the ability to perform task scheduling effectively is a measure of success in any multiprocessing environment such as a cloud deployment or job-shop scheduling or assembly line task scheduling and so on. Cost effective resource scheduling or CERS algorithm has been de- facto for many cloud deployments over the years. It is so due to the fact that the algorithm is computationally simple, has minimum overheads and is based on keeping the cloud utilization to the most optimum level. But maintaining and effective task response time it one of the main drawbacks of the CERS algorithm. In this paper, we propose a first of a kind machine learning based optimization algorithm which utilizes the same base concept as proposed by CERS, and improves it further using a combination of pre-learning and continuous adaptation techniques in order to reduce the mean response time for a given set of tasks. The proposed algorithm will further be compared with the standard CERS implementation, and the results will be evaluated in terms of resource cost, and mean response time.