Rule-based automatic software performance diagnosis and improvement

Abstract

Performance of a software system is the result of many interacting factors. This paper describes a rule-based framework to identify root causes of performance limits, to untangle the effects of the system configuration (such as the allocation of processors) from limits imposed by the software design, and to recommend both configuration and design improvements. The framework uses a performance model which represents (and is derived from) a UML design model, and applies transformations to the given performance model to obtain another improved one. The improvements imply configuration and design changes which can be applied to the system. This paper describes the approach and demonstrates feasibility by applying a small set of rules to the design of a web application.