Dominant quality of service fair allocation with bounded number of tasks in cloud computing systems

Abstract

Multi-resources allocation is a fundamental issue in cloud computing systems and most relevant researches on this topic aim at higher resources utilisation rate during allocation. Throughout our paper, we focus on cases where a series of bounded number tasks are being scheduled and seek a fair allocation among those tasks in terms of quality of service (QoS). Thus, we propose the dominant quality of service fairness (DQSF) mechanism, derived from max-min fairness algorithm and dominant resource fairness (DRF), using the total response time to measure the quality of service. DQSF is proved satisfying some properties of optimal allocation mechanism such as pareto optimality (PO), envy-freeness (EF) and others. Common extensions of DQSF are also discussed. Finally, simulations are performed to demonstrate the great improvement in dominant quality of service of the proposed mechanism, compared to DRF and slot-based fair scheduler.