Newton: Computer system improvement through comparative macroperformance analysis
Kipp, Lyle Dean
This item is only available for download by members of the University of Illinois community. Students, faculty, and staff at the U of I may log in with your NetID and password to view the item. If you are trying to access an Illinois-restricted dissertation or thesis, you can request a copy through your library's Inter-Library Loan office or purchase a copy directly from ProQuest.
Permalink
https://hdl.handle.net/2142/21045
Description
Title
Newton: Computer system improvement through comparative macroperformance analysis
Author(s)
Kipp, Lyle Dean
Issue Date
1994
Doctoral Committee Chair(s)
Kuck, David J.
Department of Study
Electrical and Computer Engineering
Discipline
Electrical and Computer Engineering
Degree Granting Institution
University of Illinois at Urbana-Champaign
Degree Name
Ph.D.
Degree Level
Dissertation
Keyword(s)
Engineering, Electronics and Electrical
Computer Science
Language
eng
Abstract
In recent years, computer systems with complex architectural features have been designed and built without a complete understanding of the performance impact of the features. All current systems with complex architectures deliver only a small fraction of their theoretical peak performance on all but the most ideal problems. We believe comparative system analysis and design can greatly enhance the process of computer system macroperformance improvement by allowing each system's designers to focus on those aspects in which their system is substantially weaker than other systems.
In this thesis, we present Newton, a philosophy and methodology for comparative analysis and improvement of computer systems. The basic methodology may be broken into three phases. First, models for each of multiple computer systems are developed; the models allocate measured running time to system components. Second, the original measurements or the running-time allocations for multiple systems are compared to detect and isolate performance problem. Finally, predictions are made of the performance improvement that will be realized if a discovered performance problem is alleviated.
For research into this methodology, software has been written which automates many aspects of the methodology, and it has been used to build models for an Alliant FX/80, a Convex C240, and the CSRD Cedar. The models are shown to represent the systems accurately in most cases. The prototype was also used to compare pairs of these systems and to predict what performance changes would be realized if various hardware and software changes were made. Verification studies indicate that these predictions are quite accurate.
When analyzing complex systems, the amount of data to examine can easily obscure the desired information. The Newton methodology has shown itself to be an effective filter for finding aspects of each system that are truly relevant to performance by quantifying the degree to which each aspect is a bottleneck to increased performance. The philosophy of comparative analysis, such as that implemented by Newton, is clearly a valuable tool for performance evaluation and improvement, although the methodology requires further development for practical use in the design and improvement of computer systems.
Use this login method if you
don't
have an
@illinois.edu
email address.
(Oops, I do have one)
IDEALS migrated to a new platform on June 23, 2022. If you created
your account prior to this date, you will have to reset your password
using the forgot-password link below.
