Professor: C. M. Woodside, Room 4482 Mackenzie Building, tel: 520-5721, email: cmw@sce.carleton.ca
Course Objectives: To understand the fundamentals of mathematical models for performance of computer and communications systems. Approximations used by modellers to simplify their models and to speed up computations. The range of modelling techniques to be covered includes queuing networks, layered queueing, Petri nets, stochastic process algebras and task graphs. For the practitioners, this course describes useful tools and when they can safely used. For the researcher, it describes the foundations of open areas of new research, and gives an introduction to the research literature on performance.
Overview:
Markov chains and queues: Summary/Review
Stochastic Petri net models and the analysis. Models of protocols and multiprocessors. The : GreatSPN tool for solving them. Petri nets are ideal for logic-intensive models with special rules that have to be modelled, like bus arbitrators or protocols.
Mean value analysis and single queues. Priorities. Powerful approach, common problems.
Networks and queues. Standard models of computer systems (central server) and communications networks. Product form models and mean value analysis. The QNAP tool for solving queueing models.
Decomposition, Norton's theorem, approximate decompositions and rate-equivalent servers. Applications to modelling subsystems (such as communications subnetworks) and resources (e.g. contention for memory).
Extended queueing networks. Surrogate delay approximations. Solutions methods for layered queueing models. The layered queueing network solver.
Other Topics: stochastic process algebras (have a lot in common with Petri nets but a structure with composition of subsystems).
Text:
D.A. Menasce et al, "Capacity Planning and Performance Modelling", Prentice Hall
Reference material on Petri Nets, etc., will be distributed.
Evaluation: There will be:
--two or three assignments on building and solving models and on the fundamentals underlying the models, --value 30%
--an oral presentation of a paper selected from the recent modelling literature ---value 10%
--an assignment with a written report and an oral presentation, to comment on a method from a paper (probably the same one), by applying it oneself --- 30%
--a short examination on terminology and concepts --- value 30%