Date: Sat, 19 Mar 2011 18:19:45 -0700 From: Navdeep Parhar <nparhar@gmail.com> To: Gordon Tetlow <gordon@tetlows.org> Cc: arch@freebsd.org, George Neville-Neil <gnn@neville-neil.com> Subject: Re: Updating our TCP and socket sysctl values... Message-ID: <AANLkTi=QRBj-u3rPaw37bLJ9yF4Qv1B11PQP-ybjkg16@mail.gmail.com> In-Reply-To: <AANLkTimOaVwy-ne6JWtaHqjtN3v6fEwbnasaRVi8M7gW@mail.gmail.com> References: <132388F1-44D9-45C9-AE05-1799A7A2DCD9@neville-neil.com> <AANLkTi=ptv617t0KhgNrcxTUzLmQd0eLFBf2x4%2BP7EAL@mail.gmail.com> <AANLkTimOaVwy-ne6JWtaHqjtN3v6fEwbnasaRVi8M7gW@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, Mar 19, 2011 at 5:38 PM, Gordon Tetlow <gordon@tetlows.org> wrote: > On Sat, Mar 19, 2011 at 4:13 PM, Navdeep Parhar <nparhar@gmail.com> wrote= : >> 256KB seems adequate for 10G (as long as the consumer can keep >> draining the socket rcv buffer fast enough). =A0If you consider 2 x >> bandwidth delay product to be a reasonable socket buffer size then >> 256K allows for 10G networks with ~100ms delays. =A0Normally the delay >> is _way_ less than this for 10G and even 256K may be an overkill (but >> this is ok, the kernel has tcp_do_autorcvbuf on by default) > > The BDP for a 10Gbps link with 100ms delay is about 120MB. I meant 100us (microseconds), sorry. My point still stands - 10G networks have much less one way delay than this. The worst I can find in the lab right now has around ~30us delay. A socket rcv bufsize of 64K maxes out the link in some casual testing with netperf (with autosizing disabled). 256K is already more than what's needed. Regards, Navdeep > > Here's a decent calculator for figuring it out: > http://www.speedguide.net/bdp.php > > Regards, > Gordon >
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?AANLkTi=QRBj-u3rPaw37bLJ9yF4Qv1B11PQP-ybjkg16>