Performance Analysis
Lecturer s : Peter Harrison (homepage) , Giuliano Casale (homepage)
For course notes click on the lecturers homepages.To introduce analytical modelling techniques for predicting computer system performance.
Motivation and survey; the need for performance prediction in optimisation and system design.
Basic probability theory: renewal processes; Markov processes; birth and death processes; the single server queue; Little's law; embedded Markov chain; M/G/1 queue; queues with priorities; queueing networks - open, closed, multi-class; equilibrium state space probabilities, proof for single class; normalising constants; computation of performance measures; convolution algorithm; mean value analysis; application to multi-access systems with thrashing.
Decomposition and aggregation: Norton's theorem; M/M/n queue; multiple independent parallel servers.
The course also offers an introduction to performance modelling using a stochastic process algebra, eg PEPA. To include: expansion law, apparent rate, steady-state
analysis, transient state analysis through uniformisation and reward
vectors.
Main campus address:
Imperial College London, South Kensington Campus, London SW7 2AZ, tel: +44 (0)20 7589 5111
Campus maps and information | About this site
© Copyright 2008 Imperial College London