Adaptive Distributed Real-Time Transaction Management in Safety-Critical Systems

Abstract

We present a novel transaction model for distributed, safety-critical real-time applications. We refrain from traditional features like roll-back, or restart after preemption because they are costly and have an unpredictable effect on meeting hard deadlines. From our MELODY system we borrow various adaptability concepts and techniques about task execution (criticality, sensitivity, survivability). These are used for stepwise generating an adaptive transaction model, which explicitly utilizes the nested transaction structure. The key idea is to save as many tasks or subtransactions from executing after a transaction invocation. This is based on a novel formal concept of task/ transaction similarity, and this is the focus of work here. A comprehensive working example is presented and discussed. Current and future work is outlined, and put into perspective with respect to previous work on real-time transaction management.