Failure Prediction for the Proactive Adaptation of SOS

Towards Accurate Failure Prediction for the Proactive Adaptation of Service-oriented Systems

Andreas Metzger
Research group leader 'Software and Service Quality'
Paluno (The Ruhr Institute for Software Technology)
University of Duisburg-Essen

DEI - Building 22, Meetings Room
July 18th, 2011
2.00 pm

Furnishing service-oriented systems with self-adaptation capabilities allows those systems to become resilient against failures of their constituent services. Especially proactive adaptation capabilities, which apply countermeasures to prevent the occurrence of failures, provide significant benefits, such as avoiding costly compensation and repair activities. An important challenge is to trigger proactive adaptations accurately; firstly, because the execution of an unnecessary proactive adaptation can lead to additional costs or failures that would not have occurred if the system would have remained unchanged; secondly,
because a missed proactive adaptation opportunity would diminish the benefits or proactive adaptation.

The presentation will discuss two directions along which accurate proactive adaptations can be achieved: (i) by improving the quality prediction techniques that trigger the adaptations (i.e., during design time); (ii) by dynamically estimating the accuracy of the predicted failures during the operation of the service-oriented system (i.e., during run-time). The discussion is backed by concrete examples of existing prediction techniques (developed in S-Cube, the EU network of excellence on software services and systems) and supported by experimental results.

Elisabetta Di Nitto

Research areas:
Advanced software architectures and methodologies