Effects of resource utilization monitoring in fault recovery

We develop a technique to reduce the time to perform online tests and diagnose faulty units and reduce the recovery time when a fault occurs in a system. A clue is given to the tester about possible faulty locations. Thus only a fraction of the resources within a system needs to be tested. This is accomplished by keeping track of the resources used by an application program when it executes. We demonstrate that a significant reduction in test time can be achieved, in particular for cache and memory subsystems. This technique can improve response time and meet more deadlines in soft real-time systems when the system employs online tests and recovery schemes. We develop this technique further and support our analysis using trace-driven simulation. We discuss ways to implement the resource utilization vector (RUV) scheme in a system, and show how the RUV scheme is used to improve the forward error recovery process.<>