الفهرس 
Introduction
Arcs (requests)Only 14 pages are availabe for public view
 |  | from | 127 |  |  |
| | | from | 127 | | |
Abstract
The quality of many software systems is determined by their performance characteristics, such as throughput and resources utilization. Developers of such systems should be able to understand performance effects on the systems at an early stage when changes are easy and less expensive. This can be achieved by constructing and analyzing performance models. Each of automation degree of evaluation process, standardization degree, and performance parameters evaluated classifies each evaluation technique. Although these are the most important characteristics of each technique, they represent a great challenge for Software Performance Engineering (SPE). This thesis presents a transformation framework to overcome these challenges. This framework transforms Unified Modeling Language (UML) software models into Layered Queuing Networks (LQN) performance models. The framework steps can be applied on any type of software especially on distributed object applications (Web Services). The framework major characteristics are: The ability to accept UML models built with more than one type of diagrams in the first phase, The standardization of transformation process because of applying XSLT-rules in all the framework phases, The two, newly implemented, algorithms that used to achieve high degree of automation in both transformation and performance evaluation process, Performance parameters such as throughput, utilization, and service time are considered in performance evaluation process. The framework for model transformation and performance evaluation consists of three main phases. In phase one: A CASE tool is used to build UML software model. In phase two: Transformation process is done in three steps. In phase three: LQN model is solved using analytical solver to evaluate performance readings. The framework steps are applied on a case study using a proposed application. The performance readings are recorded and compared with real system performance readings in order to check the validity of the proposed framework. The framework readings gave a small error ratio compared with other methods.