USDP分析设计中基于排队网络模型的软件性能预测方法  

Queuing Network Model-based Software Performance Prediction During Analysis and Design in USDP

作　　者：徐忠富[1] 陈永光[1] 杨建华[2] 彭珲[1] 

机构地区：[1]中国人民解放军63880部队 [2]国防科技大学电子科学与工程学院

出　　处：《计算机科学》2007年第10期272-277,共6页Computer Science

摘　　要：在统一软件开发过程(USDP)中,基于统一建模语言(UML)的模型是面向软件功能实现的。为了在软件开发的早期预测软件性能,基于UML的软件模型需要被扩展,增加获取和描述软件性能预测模型变量所需的信息。本文提出了在USDP分析和设计阶段预测软件性能的过程框架;定义了一个排队网络模型(QNM)元模型;基于QNM元模型,扩展软件分析和设计模型,增加软件系统应用模式描述以及协同实现软件系统功能的物理计算设备的特性和行为描述;采用基于可扩展标记语言元数据交换(XMI)的模型变换技术,生成软件性能预测QNM;通过分析所生成的QNM,可以评估和比较软件体系结构和软件设计对待实现的软件系统性能特性的影响。通过示例展示了所提出方法的可应用性。UML-based software models in USDP are functionality-oriented. For the purpose of predicting software performance in the early stage of software development, they should be extended with information required for deriving and specifying variables of performance prediction models. We present a framework for predicting software performance during analysis and design in USDP. A QNM-Metamodel is defined to provide the basis for extending the analysis and design models with specifications about system usage patterns, properties and behaviors of physical computer devices that collaborate to execute the system functionality. By means of XMI-based model transformation, QNMs are generated. Analysis results of the QNMs enable the evaluation and comparison of architectural decisions and design solutions with reference to their impacts on performance properties of the software system to be implemented. A case study demonstrates the application of the proposed approach.

关 键 词：统一建模语言 统一软件开发过程 软件性能预测 排队网络模型 元模型 可扩展标记语言 基于可扩展标记 语言的元数据交换 

分 类 号：TP311.52[自动化与计算机技术—计算机软件与理论]