This preview shows pages 1–2. 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-1 1 Modelling and Simulation 1.1 Introduction This course teaches various aspects of computer-aided modelling with an emphasis on the performance evaluation of computer systems and communication networks. The perfor- mance of a system inﬂuences its design, procurement and subsequent use. Performance analysis may be carried out at any stage of that life-cycle. In general, the problem is to achieve the highest possible performance given the constraints on the system. Typical constraints might be the number of users, the bandwidth or the cost. Modelling and simulation are methods which are commonly used by performance analysts to represent constraints and optimise performance. Many of the ideas of the course will be presented by examples; as well as computer systems and communication networks, other dynamic systems such as manufacturing and transport systems will sometimes be considered. With all the examples it will be as important to consider the underlying concepts as the details of the example itself. 1.2 Models as tools The term model conjures up a variety of images. We will use it to mean something which represents a system in a form which allows us to make predictions about the behaviour of the system. How accurate these predictions are will depend on the detail we invest in the model. In some cases, when we only need an approximate or relative measure of the behaviour of the system, a very crude model will suﬃce. When the answers obtained from the model are crucial we will be prepared to expend more effort in developing our model and the representation of the system will be less abstract. During the course we will consider a range of modelling techniques from the very abstract, in which the behaviour of the system is captured by a single equation, the so-called operational laws , to the very detailed simulation models in which every aspect of the system’s behaviour can be represented. In all cases a model should be an abstraction of the system: an attempt to distill, from the mass of details that is the system itself, exactly those aspects that are essential to the system’s behaviour. Once a model has been defined through this abstraction process, it can be parameterised to reﬂect any of the alternatives under study, and then evaluated to determine its behaviour under this alternative. Using a model to investigate system behaviour is less laborious and more ﬂexible than direct experimentation, because the model is an abstraction that avoids unnecessary detail. There are many reasons why investigations into the behaviour of a system cannot be carried out by direct experimentation on the system. At best experimentation will prob- ably be disruptive, at worst, dangerous. For example, a system manager of a heavily loaded file server is unlikely to allow experimentation, or even simple monitoring, because the disruption to users will be too great. Systems such as nuclear reactor control sys- tems or the London Underground control system are simply not safe to experiment with.tems or the London Underground control system are simply not safe to experiment with....
View Full Document
- Spring '10