Drexel University Home Pagewww.drexel.edu DREXEL UNIVERSITY LIBRARIES HOMEPAGE >>
iDEA DREXEL ARCHIVES >>

iDEA: Drexel E-repository and Archives > Drexel Theses and Dissertations > Drexel Theses and Dissertations > A decentralized control and optimization framework for autonomic performance management of web-server systems

Please use this identifier to cite or link to this item: http://hdl.handle.net/1860/2643

Title: A decentralized control and optimization framework for autonomic performance management of web-server systems
Authors: Wang, Mianyu
Keywords: Electric engineering;Electronic data processing--Distributed processing;Autonomic computing
Issue Date: 29-Jan-2008
Abstract: Web-based services such as online banking and e-commerce are often hosted on distributed computing systems comprising heterogeneous and networked servers in a data-center setting. To operate such systems efficiently while satisfying stringent quality-of-service (QoS) requirements, multiple performance-related parameters must be dynamically tuned to track changing operating conditions. For example, the workload to be processed may be time varying and hardware/software resources may fail during system operation. To cope with their growing scale and complexity, such computing systems must become largely autonomic, capable of being managed with minimal human intervention. This study develops a distributed cooperative-control framework using concepts from optimal control theory and hybrid dynamical systems to adaptively manage the performance of computer clusters operating in dynamic and uncertain environments. As case studies, we focus on power management and dynamic resource provisioning problems in such clusters. First, we apply the control framework to minimize the power consumed by a server cluster under a time-varying workload. The overall power-management problem is decomposed into smaller sub-problems and solved in cooperative fashion by individual controllers on each server. This approach allows for the scalable control of large computing systems. The control framework also adapts to controller failures and allows for the dynamic addition and removal of controllers during system operation. We validate the proposed approach using a discrete-event simulator with real-world workload traces, and our results indicate that the controllers achieve a 55% reduction in power consumption when compared to an uncontrolled system in which each server operates at its maximum frequency at all times. We then develop a distributed resource provisioning framework to achieve di┬«erentiated QoS among multiple online services using concepts from hybrid control. We use a discrete hybrid automaton to model the operation of the computing cluster. The resource provisioning problem combining both QoS control and power management is then solved using a decentralized model predictive controller to maximize the operating profits generated by the cluster according to a specified service level agreement. Simulation results indicate that the controller generates 27% additional profit when compared to an uncontrolled system.
URI: http://hdl.handle.net/1860/2643
Appears in Collections:Drexel Theses and Dissertations

Files in This Item:

File Description SizeFormat
Wang_Mianyu.pdf1.39 MBAdobe PDFView/Open
View Statistics

Items in iDEA are protected by copyright, with all rights reserved, unless otherwise indicated.

 

Valid XHTML 1.0! iDEA Software Copyright © 2002-2010  Duraspace - Feedback