IMDEA Networks Institute Publications Repository

Competitive Analysis of Task Scheduling Algorithms on a Fault-Prone Machine and the Impact of Resource Augmentation

Fernández Anta, Antonio and Georgiou, Chryssis and Kowalski, Dariusz R. and Zavou, Elli (2015) Competitive Analysis of Task Scheduling Algorithms on a Fault-Prone Machine and the Impact of Resource Augmentation. In: Workshop on Adaptive Resource Management and Scheduling for Cloud Computing (ARMS-CC 2015), 20 July 2015, San Sebastian, Spain.

[img]
Preview
PDF (Competitive Analysis of Task Scheduling Algorithms on a Fault-Prone Machine and the Impact of Resource Augmentation) - Published Version
Download (320Kb) | Preview

Abstract

Reliable task execution in machines that are prone to unpredictable crashes and restarts is challenging and of high importance. However, not much work exists on the worst case analysis of such systems. In this paper, we analyze the fault-tolerant properties of four popular scheduling algorithms: Longest In System (LIS), Shortest In System (SIS), Largest Processing Time (LPT) and Shortest Processing Time (SPT), under worst case scenarios on a fault-prone machine. We use three metrics for the evaluation and comparison of their competitive performance, namely, completed time, pending time and latency. We also investigate the effect of resource augmentation in their performance, by increasing the speed of the machine. To do so, we compare the behavior of the algorithms for different speed intervals and show that between LIS, SIS and SPT there is no clear winner with respect to all the three considered metrics, while LPT is not better than SPT.

Item Type: Conference or Workshop Papers (Invited Paper)
Uncontrolled Keywords: Scheduling, Online Algorithms, Different Task Processing Times, Failures, Competitive Analysis, Resource Augmentation.
Subjects: UNSPECIFIED
Divisions: UNSPECIFIED
Depositing User: Elli Zavou
Date Deposited: 12 Jun 2015 10:51
Last Modified: 12 Jun 2015 10:51
URI: http://eprints.networks.imdea.org/id/eprint/1069

Actions (login required)

View Item View Item