Author

Tan Fong Ang

Date of Award

1-1-2011

Thesis Type

phd

Document Type

Thesis

Divisions

deptfsktm3

Department

Dept of Computer System & Technology

Institution

University of Malaya

Abstract

The use of Grid and Cloud Computing for resource sharing has received tremendous attention in recent years. The merging of virtualization, utility computing and distributed computing with Service Oriented Architecture (SOA) has provided greater flexibility and ushered in a new paradigm of on-demand services. However, job scheduling remains a formidable challenge due to the dynamic and heterogeneous nature of Grid and Cloud Computing. Furthermore, the increasing and diverse end users requests raise new and greater urgencies to resolve the provisioning of Quality of Service (QoS). This thesis proposes a Hybrid Scheduling Algorithm (HSA) with automatic deployment mechanism that maximizes resources utilization and minimizes total makespan. Subsequently, an Adaptive Scheduling Algorithm (ASA) that uses benchmarking is proposed to enhance the HSA. ASA is able to optimize the job scheduling performance over other approaches. Finally, Adaptive QoS Scheduling Algorithm (AQoSSA), an enhancement of ASA, is presented to meet the varied users QoS requirements. AQoSSA is able to maximize reliability and profit while guaranteeing the users’ QoS requirements. An experimental testbed is developed to evaluate the performances of all the proposed algorithms. The makespan results showed that the HSA and ASA outperformed the conventional MIN-MIN and MAX-MIN by between 1% - 10% and 5% - 17% respectively. Whereas, AQoSSA outperformed MIN-MIN QoS and MAX-MIN QoS by between 3% - 6% and 10% - 47% in terms of reliability and profit respectively while guaranteeing the users’ QoS requirements.

Note

Thesis submitted in fulfillment of the requirement for the degree of Doctor of Philosophy

Share

COinS