Integrity Protocols for Recovering from Distributable Real-Time Thread Failures with Assured Timeliness in Dynamic Systems

We consider the problem of recovering from failures of distributable threads with assured timeliness in dynamic systems with overloads, and node and (permanent/transient) network failures. When a distributable thread encounters a failure that prevents its timely execution, the thread must be terminated. Thread termination involves detecting and aborting thread orphans, and delivering exceptions to the farthest, contiguous surviving thread segment for possible execution resumption. Thread termination operations must optimize system-wide timeliness. We present a scheduling algorithm called HUA and two thread integrity protocols called D-TPR and W-TPR. We show that they bound the orphan cleanup and recovery time with bounded loss of the best-effort property—i.e., high importance threads are always favored over low importance ones (for feasible completion), irrespective of thread urgency. Our implementation experience using the emerging Reference Implementation of Sun’s Distributed Real-Time Specification for Java demonstrates the algorithm/protocols’ effectiveness.