Statistical Debugging for Petascale Environments. Final Report.

摘要

Statistical debugging identifies program behaviors that are highly correlated with failures. Traditionally, this approach has been applied to desktop software. In this context, statistical debugging is effective in identifying the causes that underlie several difficult classes of bugs. These include memory corruption, non-deterministic bugs, and bugs with multiple temporally-distant triggers. The domain of scientific computing offers a new target for this type of debugging. Scientific code runs at massive scales offering massive quantities of statistical feedback data. Data collection can scale well because it requires no communication between compute nodes. Unfortunately, existing statistical debugging techniques impose run-time overhead that is unsuitable for computationally-intensive code. Additionally, the normal communication that occurs between nodes in parallel jobs violates a key assumption of statistical independence in existing statistical models. In work supported by this subcontract, we have reduce the run-time overhead of statistical debugging instrumentation by up to 20% over prior work. We have also addressed challenges related to data collection. A paper describing our work in this area has been prepared for submission to a suitable publication venue.