MA Zwischenvortrag, Lukas Egger, Ort: O27/545, Datum: 09.01.2018, Zeit: 16:30 Uhr

Interpreting the results of software performance analysis is a task which mostly relies on the knowledge of experts. Developers with less experience often struggle to make sense of the data presented to them by analysis. Therefore, many approaches try to automate the generation of feedback, usually while relying on a model of the software system. However, these models are often only approximate, if they exist at all. Especially when considering migrated legacy software, the creation of such models is usually quite difficult or just not feasible.

The subject of this master's thesis is to describe a way to automatically generate feedback suitable for performance analysis of migrated legacy software. The approach uses a formalization of performance anti-patterns. Just like patterns, anti-patterns describe recurring solutions to common problems, however, their use (or misuse) produces negative consequences. Using a formalization of these anti-patterns a framework, that is able to interpret the results of software performance analysis, can be created. The formal and structured approach allows to create reusable components, which can be combined to detect different anti-patterns. The framework allows to include domain specific expert knowledge, making it flexible and easily modifiable.