3 ms·
Why Events Are A Bad Idea (for high-concurrency servers)
- kenjackson 15y agoWorth pointing out that one of the authors, Eric Brewer, is now part of the effort to design Google's next generation infrastructure: http://twitter.com/#!/eric_brewer/status/68051541063503872 http://twitter.com/#!/eric_brewer/status/68051541063503872
- beagle3 15y agoAnd while we're at it - he is much better known for the CAP theorem: http://en.wikipedia.org/wiki/CAP_theorem http://en.wikipedia.org/wiki/CAP_theorem
- zik 15y agoI thought it was strange that in the '“Problems” with Threads' section they didn't even mention testability - which is the number 1 problem we have in trying to create high-reliability systems with threads. I work in embedded systems which run unattended for months at a time so they have to be reliable. The interactions between threads can create all sorts of unexpected race conditions which are very hard to predict. And if you can't assure repeatable behavior you can't properly test a system. On the basis of testability alone we avoid threads.
- cakoose 15y agoThis paper uses cooperatively scheduled threads (i.e. no preemption or concurrent execution of threads), which means you don't have to deal with synchronization.
- colin_jack 15y agoHe also tweeted this in relation to it: http://twitter.com/#!/eric_brewer/status/105160325824917504 http://twitter.com/#!/eric_brewer/status/105160325824917504 Its a response to http://twitter.com/#!/ericflo/status/104972704016379904 http://twitter.com/#!/ericflo/status/104972704016379904.
- wmf 15y agoWe just had this yesterday: http://news.ycombinator.com/item?id=2907415 http://news.ycombinator.com/item?id=2907415
- dblock 15y agosorry, i was hoping NH would find a dup if it were already there! :(