Adaptive Systems & Machine Learning

Self-adaptation enables a software system to execute successfully under situations that are unknown during design time. An adaptive software system thus can handle situations at run time such as the actual environment the system faces during operation, whether or not the software still has bugs, as well as when and how the requirements may change. To this end, adaptive software systems reconfigure their structure or modify their behaviour at run-time in response to their perception of themselves, their environment, and their requirements.

As an example, a cloud system during run time may face an unexpected change in its workload, such as a radical increase in the number of users that simultaneously access the system. As a consequence, the cloud system in its current configuration is not able to handle such workload and thus cannot satisfy its response time requirements. To handle this situation, the cloud systems thus may dynamically add additional compute resources to handle the increased workload.

Research Topics

  • Continuous Delivery (DevOps) and evolution of self-adaptive systems
  • Online machine learning for self-adaptive systems
  • Coordinated adaptation among cloud applications and infrastructure
  • Data protection monitoring and data-driven adaptation of cloud systems

Head of Adaptive Systems and Big Data Applications

Dr. Andreas Metzger

Software Systems Engineering (SSE)

Universität Duisburg-Essen
Gerlingstraße 16

45127 Essen
Germany
Mehr Informationen

Latest news from this area

Security for Personal Data in the Cloud

The demand for cloud services has never been greater. This brings the issue of data protection to the fore. In the RestAssured project, scientists from the Ruhr Institute for Software Technology paluno at the University of Duisburg-Essen have researched how cloud providers can better protect the personal data of their customers.

Proactive Adaption of Business Processes via Online Reinforcement Learning

Adapt or wait and see? Andreas Metzger, Tristan Kley and Alexander Palm present an AI based approach to determine when preditions are accurate enough to trigger adaptions.

Networking of European Data Platform Projects

At the beginning of the year, nine EU-funded data platform projects were launched under the umbrella of the Big Data Value PPP. The first online meetings have now taken place to establish synergies between the projects.

Further News and Press Releases