Deterministic Record-and-Replay
Research for Practice combines the resources of the ACM Digital Library, the largest collection of computer science research in the world, with the expertise of the ACM membership. Research for Practice columns have a common goal of sharing the joy and utility of reading about computer science research exchanged between academics and their counterparts in industry.
This summer's installment of Research for Practice covers a topic that, despite its maturity, continues to produce cutting-edge research: deterministic record-and-replay. Deterministic record-and-replay technologies enable a faithful re-execution (replay) of a program that ran in the past (and perhaps encountered a rare bug, a performance anomaly, or an intrusion by an adversary). But accomplishing this requires that any nondeterministic inputs to the program be logged (recorded) during execution.
The selection of techniques presented here is curated by Andrew Quinn, assistant professor of computer science and engineering at UC Santa Cruz. We chose the topic because of its growing relevance to our audience of practitioners, and we chose Professor Quinn because of his expertise in the area. His selections here all come from recent publications in top CS conferences, and they cover a range of topics from core technologies providing record-and-replay to somewhat more exotic applications. They reveal both how broadly applicable this technique is across domains and just how HOT record-and-replay remains even though it has been an active area of research for a long time.
