SAFT · Scalable autonomic fault tolerance for distributed middleware

Lines connect the coordinator with its partners. CORDIS does not always give exact coordinates for projects before 2014. These points are placed at city or country level.

Project objective

The dependable operation of the global distributed IT infrastructure is of increasing importance for our society. Tolerating failures of components is one important aspect of dependability. For a wider deployment of fault-tolerant systems, (1) the tradeoffs with other aspects of dependability, such as performance and real-time guarantees, must be better understood, and (2) systems must operate in a more autonomic manner, not requiring expert configuration and constant monitoring.This project aims at pushing the state of the art regarding these concerns, by:(1) developing fault tolerant protocols that transparently scale to a high number of nodes,(2) developing an adaptive scalable failure detection service based on a flexible abstraction,(3) investigating tradeoffs between different failure handling techniques to construct protocols with predictable behaviour in unstable environments, and(4) studying the robustness of protocols under extreme loads, which involves defining a generic framework for flow control, suitable for complex distributed protocols.All the protocols will be designed to be self-configuring and will be able to adapt to changing environmental conditions. The development process includes extensive feedback from performance studies, both simulations and measurements in clusters and wide-area networks. The project aims at realizing its goals in the context of a middleware developed at the host institute. The project will create synergies between fault-tolerant distributed computing, middleware and autonomic computing. It will enable the fellow to become an expert in the two latter fields, improve his teaching and research management skills, and aid his reintegration after a stay in Japan. Collaboration with German, Swiss and Japanese institutes is foreseen, and the contact with three different communities would benefit the fellow's future academic career.

Original text from CORDIS.

Participants

• UNIVERSITäT ULMCoordinatorCity levelGermany

Links

• View on CORDIS