This preview shows pages 1–3. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.View Full Document
Unformatted text preview: CS4 Modelling and Simulation LN-17 17 Comparison of Techniques 17.1 Introduction In this note we review the various approaches to representing a system which we have considered during the course, and try to identify their relative strengths and weaknesses. In doing so we gain some insight into when it is appropriate to use each of the techniques. During the course the following ways of representing the performance related aspects of the behaviour of computer and communication systems have been discussed: • Operational laws • Markov processes, and related high level modelling paradigms: – Stochastic Petri Nets – Stochastic Process Algebras – Queues and Queueing Networks • Simulation • Measurement The criteria on which we will compare the different approaches are the following: • Time and skill requirements • Expressiveness/assumptions • Diagrammatic representation • Solution • Deriving performance measures • Particular strengths or weaknesses However, you should note that not all these criteria are applicable to all the approaches, and so they will not all be considered in every case. 17.2 Operational laws Time and skill requirements The key advantage of operational laws is their minimal requirements in terms of time and skill. This makes them especially useful for calculating bounds on performance measures rather than specific values. For example, worst and best timing assumptions can be quickly evaluated to give an estimate of the range of a performance measure for a system. The resulting bounds can contribute significantly to the analyst’s understanding of the system. The calculations involved are simple, and can often even be done by hand. Expressiveness/assumptions The assumptions which are necessary for the applica- tion of the operational laws are that the system is job ﬂow balanced ( conservation of work ), and that the workload and the system are homogeneous . The extremely abstract view taken of the system means that the expressiveness is poor: many aspects of system behaviour cannot be represented. For example, it is impossible to include any notion of relative ordering between jobs, synchronisation, or exclusive access. This can be viewed both as a strength and as a weakness. 119 CS4 Modelling and Simulation LN-17 Solution and deriving performance measures The calculations involved are simple, and can often even be done by hand and unlike other techniques there is no distinction between solving the model and deriving performance measures. Particular strengths or weaknesses The main strength of the operational laws is the speed and ease with which they can be formulated and evaluated. 17.3 Markov Processes Time and skill requirements Modelling directly at the level of the Markov process is time consuming and error-prone for all but the simplest models, even with software support. However no sophisticated modelling techniques need to be learned and the solu- tion procedure is a straightforward application of linear algebra, once the global balance...
View Full Document
- Spring '10