5

Following What is the most famous computer bug of all time?, what are the most famous concurrency-related bugs?

"Famous" means that the issue enjoyed coverage in mass media. Embarassingly costly issues, issues resulting in loss of life, or issues having an impact on politics are also considered.

"Concurrency-related" means that the issue was related to races, nondeterministic program outcome, or thread interaction in general.

E.g., the Therac-25 incidents and the Northeast Blackout of 2003 (@Kevin McKenzie: thx!) are famous and concurrency-related. But they are old, so, newer failures are appreciated.

There are some general famous-bug enumerations (e.g., http://en.wikipedia.org/wiki/List_of_software_bugs), but they are difficult to filter for concurrency issues.

  • Out of curiosity, why are you specifically looking for concurrency related bugs? – Yu Zhang Jun 15 '17 at 22:49
4

There are two that come to mind: Therac-25 and the Northeast Blackout of 2003.

The Therac-25 was a radiation therapy machine; there was a race condition that could cause the high-power electron beam to trigger instead of the low power beam. Three people died as a result. When I was in college, the Therac-25 bug was brought up when we were learning about concurrency to illustrate the dangers, as it had lead to deaths, which was at the time unusual for software bugs.

However, that was prior to the Northeast Blackout; while there were a number of things that had to happen in order for the bug to be exposed, eventually, it was the race condition that prevented an alarm from going off, which meant that the operators didn't know they needed to do anything to redistribute power, which eventually cascaded into a power failure for much of Northeast United States that lasted for several days. I'd imagine this is what's talked about in college today, since some students probably remember it.

1

There were many bugs with the initial rollout of healthcare.gov, and definitely concurrency played a role. According to Todd Park, the site was designed to handle 50K concurrent users instead of the 250k that actually tried to use it. On the first day of operation, only 6 people were able to complete the signup.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy