ADAPTATION - Algorithms to ADAPTive FAulT MonItOriNg and their implementation on CORBA

This paper presents ADAPTATION-(A) under bar lgorithms to a (DAPT) under bar ive f (A) under bar ul (T) under bar mon (I) under bart (O) under bar ri (N) under barg for asynchronous distributed systems and their implementation on CORBA. Our algorithms vary the timeouts based on a recent history of last elapsed times of the monitoring messages. The aim of the proposed algorithms is to provide a better response time to crashes and a minimum discrepancy between a suspection due to the network overload and due to the real process crash. The proposed approach extends the Fault Tolerant CORBA OMG specification with the push model and the definition of pull and push ADAPTATION fault monitors. Some ADAPTATION experiments on ACE+TAO were made to observe their behavior on changing network workloads.