Reducing Monitoring Overhead by Integrating Event- and Time-triggered Techniques

TitleReducing Monitoring Overhead by Integrating Event- and Time-triggered Techniques
Publication TypeConference Paper
Year of Publication2013
AuthorsWu, W., D. Kumar, B. Bonakdarpour, and S. Fischmeister
Conference NameProc. of the International Conference on Runtime Verification (RV)
Date PublishedSept

Runtime verification is a formal technique used to check whether a program under inspection satisfies its specification by using a runtime monitor. Existing monitoring approaches use one of two ways for evaluating a set of logical properties: (1) event-triggered, where the program invokes the monitor when the state of the program changes, and (2) time-triggered, where the monitor periodically preempts the program and reads its state. Realizing the former is straightforward, but the runtime behaviour of event-triggered monitors are difficult to predict. Time-triggered monitoring (designed for real-time embedded systems), on the other hand, provides predictable monitoring behavior and overhead bounds at run time. Our previous work shows that time-triggered monitoring can potentially reduce the runtime overhead provided that the monitor samples the program state at a low frequency.

In this paper, we propose a hybrid method that leverages the benefits of both event- and time-triggered methods to reduce the overall monitoring overhead. We formulate an optimization problem, whose solution is a set of instrumentation instructions that switches between event-triggered and time-triggered modes of monitoring at run time; the solution may indicate the use of exactly one mode or a combination of the two modes. We fully implemented this method to produce instrumentation schemes for C programs that run on an ARM Cortex-M3 processor, and experimental results validate the effectiveness of this approach.

Refereed DesignationRefereed
Related files: 


Looking for motivated students (undergrads and grads) interested in working on embedded software and systems research. Mail Sebastian Fischmeister for further information.