Date of Award
2006
Degree Type
Thesis
Degree Name
Master of Engineering Science
Program
Electrical and Computer Engineering
Supervisor
Dr. Miriam A. M. Capretz
Abstract
In today’s inter-networked and information savvy world, developers of large-scale applications are experiencing an increase in complexity of their software systems due to the continual integration of new services, software, and hardware. The concept of multithreading and, consequently, multi-threaded software, have become the norm, with most software applications today consisting of multiple threads to achieve improved performance and concurrency. However, this performance gain comes at the cost of an increase in the complexity of applications to such an extent that it has begun to exceed the human ability to design, manage, and secure these applications. This complexity affects not only the developers but also the administrators who manage the software systems and the hardware. The recognition of this problem has instigated various research efforts for simplifying and easing the management and maintenance of multithreaded software. This thesis presents one such solution to address this problem of increased maintenance complexity in the form of a novel approach to autonomously balance thread loads. The two primary design paradigms used in this research include autonomie computing, which is a vision that strives for system self-management by making quick and smart decisions, and fuzzy logic, which provides the ability to analyze multiple crisp inputs to make a single crisp output based on a set of pre-defined rules. This research integrates the best features of both paradigms and the resultant simulation exhibits properties such as self-optimization where in the fuzzy logic reasoning engine determines the appropriateness of existence and execution of a thread on a machine, and selfconfiguration, where inappropriate threads are relocated to other systems for improved performance. The simulation has been implemented in Java 1.5 and a case for the validation of the proposed approach is presented through a detailed discussion of three scenarios covering a range of system and network conditions, along with detailed results presented as evidence for the feasibility of the approach.
Recommended Citation
Baker, Jack F.G., "Autonomic Thread Load Balancing Using Fuzzy Logic" (2006). Digitized Theses. 4718.
https://ir.lib.uwo.ca/digitizedtheses/4718