Date: Mon, 15 Jul 2013 13:31:35 -0700 From: Adrian Chadd <adrian@freebsd.org> To: trafdev <trafdev@mail.ru> Cc: Sepherosa Ziehau <sepherosa@gmail.com>, freebsd-net@freebsd.org Subject: Re: SO_REUSEPORT: strange kernel balancer behaviour Message-ID: <CAJ-VmomNCvyqWoXshNPCF_yBt_hg7xmwW_GWMBcCJCki8RXAug@mail.gmail.com> In-Reply-To: <51E455D5.2090403@mail.ru> References: <51E0E2AF.7090404@mail.ru> <CAMOc5cz6gP2N62T4QhbTdVar94O4FSdPDsqktD_9vJ0mYVqt_Q@mail.gmail.com> <51E44E2F.8060700@mail.ru> <CAJ-VmomHHfhExa4g63tT_sf0hTPa2T7jPKQGHrD0fchq=-k%2B=g@mail.gmail.com> <51E455D5.2090403@mail.ru>
next in thread | previous in thread | raw e-mail | index | archive | help
On 15 July 2013 13:04, trafdev <trafdev@mail.ru> wrote: > Yep I think it's wasting of resources, poll manager should somehow be > configured to update only one process/thread. > Anyone know how to do that? > Thanks. Well, the problem here is deciding which thread to throw the request at. If the threads are equally busy, you're ok. If the threads aren't equally busy, then just doing a round robin or hash may end up keeping one thread more busy than others, skewing the workload. I'm also interested in the whole CPU pinning of sockets to line it all up with multi-queue network cards and such. Hence why I'm starting down this path. -adrian
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAJ-VmomNCvyqWoXshNPCF_yBt_hg7xmwW_GWMBcCJCki8RXAug>