Friday, April 11, 2008
Another blast from the past
I was in Neuchatel this week for some meetings and one of our conversations moved on to failure detection/failure suspecting: the fact that you cannot reliably detect failures until (and unless) those failures are eventually recovered from. Typical "detection" uses timeouts and if you use the wrong value you can end up in a world of pain. That's where failure suspectors come in: the idea is that if you think something has failed then you make sure everyone else agrees with you so even if you are wrong you don't end up with split-brain syndrome. This reminded me of some work I did back in the 90's around quantum mechanics and failure detectors.