Lessons Learned on Assumptions and Scalability with Time-Aware Instrumentation

Software instrumentation is a key technique in many stages of the development process. Instrumentation is particularly important for profiling, debugging, performance evaluation, and security analysis of real-time and embedded systems. Unfortunately, typical software-based instrumentation methods, while useful to extract high-level information from programs, concentrate on preserving only logical correctness and are thus inadequate for time-sensitive applications for which timing must also be preserved.  Time-aware instrumentation is a new view on code instrumentation, one that considers extra-functional properties and specifically timing constraints of instrumented programs. Time-aware instrumentation enables instrumenting software systems while still guaranteeing their timing requirements. This paper summarizes the work on time-aware instrumentation and highlights the lessons learned on assumptions and scalability. Specifically, it shows how strict assumptions enable a strong formal model at the expense of applicability. Subsequent relaxing of assumptions then permits scaling the concept and applying it to large real-word applications with millions lines of code. We believe that these lessons may help steer other research projects in the systems area.