Traceability Based Dynamic Metrics as a Tool for Evaluating Partitioning Decisions in Object-Oriented Architecture



This research investigates a new method of assessing predicted performance characteristics of Object-Oriented software architectures. As scenarios represent the anticipated runtime behavior of a system, they can be used at a very early stage in the design to quantitatively predict aspects of that behavior. The proposed approach can be applied earlier than other existing methods such as simulations and executable designs, and does not require special analytical skills. Early assessment is an important risk mitigation technique, that can contribute significantly to the task of building a distributed system with a greater degree of confidence in its ability to perform as required. The research involves formally defining a set of scenario-based metrics and a corresponding system-wide dynamic metrics model. A strongly typed traceability scheme is defined to support the definition and automated querying of relationships between performance requirements, scenarios, and architectural components. Scenario invocation rates are specified within the context of typical usage patterns, and the relationship between scenarios is investigated to determine when performance requirements should be propagated from one scenario to another. A supporting tool is developed and used to capture metrics and validate them against runtime and simulation measurements taken from a variety of distributed systems.

2002-08-31 to 2004-05-31
Award Amount: 
Award Number: