Allgemeine Informationen

Veranstaltungsart

Vorlesung und Übung

Semester

WiSe 2025/26

ECTS-Punkte

6

Veranstaltungsnummer

6.636

Voraussetzungen

Practical assignments will use Python3; basic Python knowledge is recommended.

Details

Link zur Veranstaltung in StudIP

Dozent*innen

• Dr. rer. nat. Birte Friesel

Beschreibung

Many of today's software systems offer a wide range of configuration options. For instance, database management systems expose a variety of global, per-connection, and per-query tunables, and operating system kernels such as Linux come with tens of thousands of configurable features. These features do not just affect functional system properties (e.g., the selection of a certain scheduling algorithm in Linux), but also non-functional properties such as kernel size, database query latency, or scheduler fairness/throughput. Considering the sheer amount of configuration options, describing their influence on non-functional system properties manually is, in general, out of the question.

This lecture covers the automatic extraction of non-functional system properties (benchmarking), machine learning methods for transforming those into functions that predict non-functional system properties from system configurations (modelling), and using the resulting models to understand and improve system performance (optimization). After attending it, students should be capable of acquiring benchmark data from software projects, building performance models for arbitrary configurable systems, and interpreting the resulting models to gain knowledge about system performance.

Dealing with non-functional system properties is an interdisciplinary affair. As such, this lecture touches several domains of (or related to) computer science, including software engineering, machine learning, statistics, theoretical computer science, and physics / electrical engineering. In all cases, basic knowledge (as taught by the Bachelor programme's mandatory modules) is sufficient.

Lecture materials and assignments will be provided in English, and assignments may be handed in in German or English. Depending on the enrolled students, the leture itself will be given either in English or German.

Studienbereiche

• Informatik > Master of Science in Informatik
• Informatik > Vorlesungen

Modulzuordnungen der Veranstaltung

• Master of Science Informatik > INF-INF-ME-6-O - Benchmarking, Modeling and Optimization of Non-Functional System Properties
• Master of Science Umweltsysteme und Ressourcenmanagement > INF-INF-ME-6-O - Benchmarking, Modeling and Optimization of Non-Functional System Properties