Resource allocation Virtual computer systems Computer Engineering
The ability to dynamically allocate system resources in a large scale distributed system is highly desirable. Dynamically allocating system resources can significantly reduce under-utilization of system resources and reduce the power consumed by the servers. Since typical enterprise computing systems consist of hundreds of servers, it is almost impossible to manually reconfigure each system parameter for optimal performance. Prior work has shown that by posing the dynamic resource provisioning problem as one of sequential optimization, we can dynamically allocate system resources for optimal performance in a dynamic operating environment. However, a single threaded implementation of this control technique does not scale well with increasing system size. Therefore, this thesis develops a parallel controller implementation for dynamic resource allocation using the OpenMP interface. We analyze the performance of this controller in a virtualized computing environment, and show that dynamic resource allocation can lead to an average of 30% savings in energy consumption, over an uncontrolled system. Parallelizing the controller also significantly reduces its execution time overhead, by as much as 263%, a compared to single threaded implementation.
Metrics
26 File views/ downloads
12 Record Views
Details
Title
A parallel controller implementation for dynamic resource allocation in virtualized computing environment
Creators
Shahab Ahmad - DU
Contributors
Nagarajan Kandasamy (Advisor) - Drexel University (1970-)
Awarding Institution
Drexel University
Degree Awarded
Master of Science (M.S.)
Publisher
Drexel University; Philadelphia, Pennsylvania
Resource Type
Thesis
Language
English
Academic Unit
College of Engineering (1970-2026); Electrical (and Computer) Engineering [Historical]; Drexel University