Bug isolation via remote program sampling

TLDR

The authors propose a low‑overhead sampling infrastructure to collect execution data from a program’s user community. The infrastructure transforms assertion‑dense code to share assertion costs, uses broad predicate guesses with elimination when assertions are absent, and applies logistic‑regression modeling to correlate behaviors with failures, enabling bug isolation. Sampled instrumentation is shown to isolate bugs, including non‑deterministic memory corruption via logistic‑regression‑based behavior correlation.

Abstract

We propose a low-overhead sampling infrastructure for gathering information from the executions experienced by a program's user community. Several example applications illustrate ways to use sampled instrumentation to isolate bugs. Assertion-dense code can be transformed to share the cost of assertions among many users. Lacking assertions, broad guesses can be made about predicates that predict program errors and a process of elimination used to whittle these down to the true bug. Finally, even for non-deterministic bugs such as memory corruption, statistical modeling based on logistic regression allows us to identify program behaviors that are strongly correlated with failure and are therefore likely places to look for the error.