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 LN14 14 Model Validation and Verification 14.1 Introduction Whatever modelling paradigm or solution technique is being used, the performance mea sures extracted from a model will only have some bearing on the real system represented if the model is a good representation of the system. Of course, what constitutes a good model is subjective, but from a performance modelling point of view our criteria for judg ing the goodness of models will be based on how accurately measures extracted from the model correspond to the measures which would be obtained from the represented system. By its nature a model is more abstract than the system it represents (even a simulation model). Viewed in one way, abstraction, and assumptions we make to achieve it, eliminate unnecessary detail and allow us to focus on the elements within the system which are important from a performance point of view; viewed in another way, this abstraction process introduces inaccuracy. Some degree of inaccuracy may be necessary, desirable even, to make the model solution tractable and/or eﬃcient. Inevitably some assumptions must be made about the system in order to construct the model. However, having made such assumptions we must expect to put some effort into answering questions about the goodness of our model. There are two steps to judging how good a model is with respect to the system. We must ascertain whether the model implements the assumptions correctly (model verification) and whether the assumptions which have been made are reasonable with respect to the real system (model validation). We have already seen examples of both model verification and model validation in models which we have considered earlier in the course. For example, in the Markov process model of the PCLAN we carried out a walkthrough of the simplified model with just two nodes in the network to check the characterisation of state which we were using. This was model verification. The level of abstraction which we had first chosen did not allow us to represent the behaviour of the system and make the assumptions about memoryless behaviour. The walkthrough allowed us to detect that we needed to distinguish states more finely, leading to a modification of the model. The SPNP modelling package allows us to include assertions within the program representing a model. These assertions are used to encode invariants about the behaviour of the system which we know should be true at all times. For example, in the readerwriter system we were able to assert that it should never be the case that a reader and a writer had access to the database at the same time. This can be used for both model verification and model validation....
View
Full
Document
This note was uploaded on 04/08/2010 for the course COMPUTER E 409232 taught by Professor Mohammadabdolahiazgomiph.d during the Spring '10 term at Islamic University.
 Spring '10
 MohammadAbdolahiAzgomiPh.D

Click to edit the document details