Abstract: A time-triggered monitor runs in parallel with the program under inspection and periodically samples the program state to evaluate a set of properties. However, a time-triggered monitor working with a fixed sampling frequency often suffers from redundant sampling, which results in excessive overhead. In this paper, we propose an effective approach to reduce redundant sampling. Our approach calculates the sampling frequency with respect to the program behavior at run time. We further advance this approach to dynamically adjust the sampling frequency at run time by predicting the program behavior using symbolic execution. Experiments show that our approach reduces the sampling frequency, runtime overhead, and the number of redundant samples by up to 3.5 times, 69%, and 86%, respectively.
Loading