[Pdns-users] multithreaded epoll_wait in PDNS
Hello Bert, I read your post regarding multithreaded epoll_wait behavior on lkml a couple of months ago: " http://lkml.org/lkml/2010/3/3/441 " It's very interesting topic to me and I would like to dig deeper; Davide seemed to purpose EPOLLET and EPOLLONESHOT as solution. However I cannot find any "edge-triggered" epoll code in multiplexer class in repository, is this issue resolved? Would you mind sharing what you've found while trying out this approach? Thank you in advance! ___ Pdns-users mailing list Pdns-users@mailman.powerdns.com http://mailman.powerdns.com/mailman/listinfo/pdns-users
Re: [Pdns-users] multithreaded epoll_wait in PDNS
On Tue, 1 Jun 2010 16:12:45 -0500, Naked Short-Selling wrote: > I read your post regarding multithreaded epoll_wait behavior on lkml a > couple of months ago: > " > http://lkml.org/lkml/2010/3/3/441 > " My understanding is that using EPOLLET would be the right approach - and particularly for UDP sockets it should be quite easy to implement it that way (it might be a bit more tricky to get the locking right for TCP sockets). Christof -- http://cmeerw.org sip:cmeerw at cmeerw.org mailto:cmeerw at cmeerw.org xmpp:cmeerw at cmeerw.org ___ Pdns-users mailing list Pdns-users@mailman.powerdns.com http://mailman.powerdns.com/mailman/listinfo/pdns-users
Re: [Pdns-users] multithreaded epoll_wait in PDNS
On Sun, Jun 06, 2010 at 06:40:18PM +0200, Christof Meerwald wrote: > On Tue, 1 Jun 2010 16:12:45 -0500, Naked Short-Selling wrote: > > I read your post regarding multithreaded epoll_wait behavior on lkml a > > couple of months ago: > > " > > http://lkml.org/lkml/2010/3/3/441 > > " > > My understanding is that using EPOLLET would be the right approach - and > particularly for UDP sockets it should be quite easy to implement it that > way (it might be a bit more tricky to get the locking right for TCP > sockets). Christof, "Naked", This issue has been studied and turns out not to be much of a problem under operational load - the number of spurious wakeups is actually minimal. We are probably saved by the fact that under high load, a new packet is probably available anyhow in most cases - perhaps not the one we were originally awoken for. One day, we'll address the epoll_wait optimization, but there are more pressing things on the horizon. Like DNSSEC. Thanks for your attention though! Bert ___ Pdns-users mailing list Pdns-users@mailman.powerdns.com http://mailman.powerdns.com/mailman/listinfo/pdns-users