Computer System Performance Evaluation (CSC5420)

Instructor: Dr. John C.S. Lui

Often times when we propose a new computer architecture, a new algorithm, a new network protocol, a new distributed database concurrency control algorithm, ...etc, we need to quantify the system performance. For example, the system response time, the expected throughput, the system stability condition, the system reliability and availability. To make these quantitative claims, we need the formal mathematical background as well as the analytical knowledge. The aim of this course is to introduce and prepare students to have this formal background, and hopefully, facilitate in your future research.

The tentative outline of the course is as follow:

• brief review of probability, combinatorics, random variables
• stochastic processes
• queueing theory such as M/M/1 and M/G/1. (Note: if time allows, we will study G/M/m, G/G/1)
• matrix-geometric solution technique
• queueing networks, product forms and various solution techniques
• matrix geometric solutions to queueing theory
• stochastic sample path analysis techniques
• simulation and variance reduction techniques
• applications of performance evaluations to computer/communication systems. For example, distributed resource allocation, computer interconnection networks, parallel computational models, communication protocol analysis, communication networks and mobile computing.

Note that:

• since the CS/CE students do not have a good background in probability and stochastic processes, I will spend some time in these topics. But students NEED TO WORK VERY HARD to keep up to the speed.

• throughout the course, I will (hopefully) motivate students by various computer related examples, for example, OS scheduling algorithms, performance modeling of I/O system, multiprocessing systems, computer networks..etc.

• I strongly encourage any final year undergraduate students to take this class, especially if you plan to go for graduate study. Hopefully, you can view this class as an advanced preparation for your graduate study. My view is that formal mathematical background is necessary if you want to explore areas in mobile computing, distributed multimedia systems, parallel/distributed processing and distributed database systems.

• Yes, the course is designed for students in the Department of Computer Science & Engineering only. Graduate students in other department are STRONGLY ENCOURAGE to take similar courses in their department. I believe students in other department are more familiar about probability, stochastic processes...etc and therefore, should be able to LEARN more by taking other courses. In another word, I don't want a student who took a similar course in another department and come to register for this course simply to get an easy grade. The purpose of graduate study is to expand our knowledge, not to get an easy grade.

• lecture time will be arranged after we have the confirmation about the registration information. Therefore, the time listed in the registration booklet may not be accurate.

p.s: I hope this course can provide a fundamental background for your research work. On the other hand, I don't think this course can help you to find a job in the Hong Kong industry. Therefore, take this course simply for the intellectual and academic interest.

Related textbooks

• Queueing Systems Volume I: Theory
  by L. Kleinrock (Wiley Interscience)

• Performance Modeling of Communication Networks and Computer Architectures
  by P.G. Harrison and N.M. Patel (Addison Wesley)

• Probability, Stochastic Processes and Queueing Theory
  by Randy Nelson, (Springer-Verlag)

• Queueing Analysis, Volume I, II and III ( reference book )
  by H. Takagi (North Holland)

• Various materials provided by the instructor

Lecture Notes