A Framework for Bottleneck Analysis in High-Performance Software Design

Abstract

The design of complex software systems is a challenging task because it involves a wide range of quality attributes such as security, performance, reliability, to name a few. Dealing with each of these attributes requires specific set of skills, which quite often, involves making various trade-offs. This paper proposes a Model-Driven Software Performance Engineering (MDSPE) process that integrates performance requirements analysis with traditional software architecture design activities. One of the most important aspects of software performance engineering is bottleneck identification and removal. We propose in this paper a new framework for performance bottleneck analysis and removal based on a simple tabular notation. A case study is presented to illustrate how the proposed framework can comply with well-known performance models to assess the performance measures and make appropriate design decisions.