الفهرس Only 14 pages are availabe for public view
 |  | from | 186 |  |  |
| | | from | 186 | | |
Abstract
Performance evaluation is required at every stage in the lik cvcle of a system. The three techniques for performance evaluation of systems are analytical modeling, simulation, and measurement. Analytical modeling is the method of choice for a fast and cost effective evaluation of systems. One of the most efficicnt high level formalisms used for analytical modeling of computers and communications systems is Petri nets (PN). Stochastic Petri nets and its variant have been used extensively in systems modeling and evaluation. Unfortunately, complexity of nowadays systems . which results in the state space explosion of the underlying Markove Chain (MC) model, has hindered the wide use of PN in the analysis of these systems.
Decomposition techniques are considered one of the most efficient methodologies
used to cope with the state explosion problem. It uses the ”divide and conquer” prin-ciple, the original large model is divided to several smaller sub n;. els results are conquered into an overall solution. Unfortunate! tion techniques either require special form of the model or give of the solution. These constraints clarify the need for an efficient technique to be used in evaluation of complex systems.
In this thesis, an efficient decomposition technique is proposed quire any special structure of Stochastic Petri Nets (SPN) mode-; high accuracy compared to other techniques introduced in the u nique overcomes most of drawbacks of other techniques propo . ,i introduces a new aggregation method that depends on using a ika called intelligent transition. The proposed technique adopts delay equivalent criteria but in a new way that differ from other techniques using the sa;i.. new technique depends on replacing parts of the system by behavior and delay equiva - lent blocks, hence it called Behavior and Delay Equivalent Blocl- i tion technique.