841: Computer Performance Evaluation Level
Dr. Jozo J. Dujmovic
(English spelling: Yozo Dooymovich)
Office: Thornton 906(MWF 11-12 a.m.)
Phone: (415) 338-1108
A systematic presentation of specific analytic,
measurement, and simulation methods for computer systems performance analysis.
The course prepares students to effectively solve computer performance analysis
problems related to the capacity planning, management, design, tuning, and
comparison of computer systems.
Prerequisites for this course are CSC 656
or ENGR 456, or consent of instructor. The course is offered to graduate
students and advanced undergraduate students.
Computer performance analysis is required
at any stage in the life cycle of a computer system, including system design,
system selection/procurement, and system use/tuning. The first problem is
to predict the values of performance indicators of a computer system during
its design and development. Then, in the area of system comparison and selection
the problem is to evaluate and compare the performance of existing competitive
systems, which are assumed to be available for performance measurements.
In the area of system management the problem is to improve the performance
level of existing operational computer systems. The main techniques for
computer performance analysis include (1) analytic modeling (queuing theory,
and operational analysis), (2) simulation (the use of specialized simulators,
and general purpose simulation languages), and (3) measurement (benchmarking
based on natural and synthetic workloads). This course includes the presentation
of all main techniques for computer performance analysis. At the beginning
individual performance indicators for all major components of the computer
hardware and software are presented. This analysis includes main memory,
processors, channels, disks, tapes, printers, compilers, and operating systems.
The central part of the course is devoted to stochastic, operational, and
simulation techniques for modeling the dynamic behavior and for computing
performance indicators of a variety of complex computer systems. The techniques
for workload characterization, benchmark design, monoprogramming performance
measurement, multiprogramming performance measurement, and interactive performance
measurement are also presented. In addition to theoretical results the course
also includes experimental computer performance analysis techniques and
laboratory experiments with modern HW/SW systems.
R. Jain, The Art of computer systems Performance
Analysis. J. Wiley, 1991.
Instructional methods include lectures, practical
laboratory work (performance measurement and modeling), homework assignments
(projects), reading and presentation of recent research papers, as well
as participating in PERNET presentations.
Homework , 20 points
Midterm exam , 30 points
Final exam , 50 points
PERNET / Extra work , 5 points.
CSc 841: Computer Performance Evaluation