Soft-set based DDQ scheduler for Optimal Task Scheduling under Uncertainty in the Cloud
Kumbham Bhargavi, B. Sathish Babu · 2017
Tasks scheduling in the cloud is one of the preliminary requirement to achieve highly optimized performance by efficient allocation of tasks to resources. But cloud paradigm is vulnerable due to the uncertainty in the task and resource parameters; hence there is a need to develop self-learning models which handle uncertainty properly before taking scheduling decisions. But in the literature, sufficient works have not been carried out to generate approximation models to take scheduling decisions under uncertainty. This paper addresses the problem of scheduling in an uncertain cloud environment, here before taking the scheduling decisions the uncertainty in the task and resource parameters is handled i.e., the task uncertainty is handled via soft-set and the resource uncertainty is managed via K-means clustering technique. After handling the uncertainty, Double-D- Q learning scheduler (DDQ) is designed to draw optimal policies using double Q values in order to efficiently map the soft-set of tasks to clusters of the resource. The performance of the soft-set enabled DDQ scheduler is analyzed using expected value method and is found to be good in terms of convergence rate and accuracy but remains moderate in terms of execution time and utilization of the resources.