Uncertainty Aware T2SS Based Dyna-Q-Learning Framework for Task Scheduling in Grid Computing

Abstract

Task scheduling is an important activity in parallel and distributed computing environment like grid because the performance depends on it. Task scheduling gets affected by behavioral and primary uncertainties. Behavioral uncertainty arises due to variability in the workload characteristics, size of data and dynamic partitioning of applications. Primary uncertainty arises due to variability in data handling capabilities, processor context switching and interplay between the computation intensive applications. In this paper behavioral uncertainty and primary uncertainty with respect to tasks and resources parameters are managed using Type-2-Soft-Set (T2SS) theory. Dyna-Q-Learning task scheduling technique is designed over the uncertainty free tasks and resource parameters. The results obtained are further validated through simulation using GridSim simulator. The performance is good based on metrics such as learning rate, accuracy, execution time and resource utilization rate.

Analysis of Energy and Network Cost Effectiveness of Scheduling Strategies in Datacentre

“…The speed, efficiency, and optimal use of resources are largely determined by the type of schedule selected for the cloud computing environments. The various scheduling criteria are maximum CPU usage and minimum throughput [22,23].…”

APVNFC: Adaptive Placement of Virtual Network Function Chains

“…Shoc (Serverless HPC over Cloud) architecture integrates scaling and scheduling policies to deal with Cloud platforms like Azure, Alibaba Cloud, Salesforce, OpenStack, and Amazon Web Services. B h a r g a v i and S h i v a [42] use the Dyna-Q-Learning task scheduling technique to manage behavioral and primary uncertainty task and resource parameters.…”