By increasing the importance of the performance in industrial and business software systems, efficient approaches to model-based performance engineering are becoming an inherent part of the development life cycle. Performance engineering at abstract levels of the software development process has an important effect on concluding the success of the software by obtaining the knowledge of optimal alternative designs. This paper introduces the performance-driven software development approach and a prediction technique that regards performance quality attributes at the abstract levels of the software development in an incremental refinement manner. The approach provides Z -based specification formalism at the meta-model level in which its instance models are automatically transformed into the formal performance analytical model, called refinable state machine (RSM). This paper analyses the throughput of a RSM by performing an approximation algorithm on two experimental case studies to determine weights of subjective performance characteristics. The approach can use the inherent performance parameters according to product usage and derive an incremental probabilistic policy determination method under design decisions in the performance plan hierarchy. The results exhibit significant support of abstract level performance profiling in terms of the throughput values.

中文翻译：

---

性能驱动的软件开发：高质量需求工程的增量细化方法

通过提高性能在工业和商业软件系统中的重要性，基于模型的性能工程的有效方法正在成为开发生命周期的固有部分。软件开发过程抽象级别的性能工程通过获得最佳替代设计的知识，对确定软件的成功具有重要影响。本文介绍了性能驱动的软件开发方法和预测技术，该技术在软件开发的抽象级别以增量细化的方式考虑性能质量属性。该方法在元模型级别提供基于 Z 的规范形式，其中其实例模型自动转换为正式的性能分析模型，称为可精炼状态机 (RSM)。本文通过对两个实验案例研究执行近似算法来确定主观性能特征的权重，从而分析 RSM 的吞吐量。该方法可以根据产品使用情况使用固有的性能参数，并在性能计划层次结构中的设计决策下导出增量概率策略确定方法。结果显示出对吞吐量值方面的抽象级性能分析的重要支持。该方法可以根据产品使用情况使用固有的性能参数，并在性能计划层次结构中的设计决策下导出增量概率策略确定方法。结果显示出对吞吐量值方面的抽象级性能分析的重要支持。该方法可以根据产品使用情况使用固有的性能参数，并在性能计划层次结构中的设计决策下导出增量概率策略确定方法。结果显示出对吞吐量值方面的抽象级性能分析的重要支持。