Date: Tue, 25 Jul 2006 11:09:43 +0100 From: "Peter Edwards" <peadar.edwards@gmail.com> To: "Jim Rees" <rees@umich.edu> Cc: fs@freebsd.org, rick@snowhite.cis.uoguelph.ca Subject: Re: freebsd4.11 patch for nfs over tcp Message-ID: <34cb7c840607250309t1b20179icb034cad7e720e7f@mail.gmail.com> In-Reply-To: <20060724152823.C32D81BB93@citi.umich.edu> References: <200607241456.KAA63411@snowhite.cis.uoguelph.ca> <20060724152823.C32D81BB93@citi.umich.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
On 7/24/06, Jim Rees <rees@umich.edu> wrote: > Delayed acks when properly implemented should not cause a performance hit. > The problem is that there are many buggy implementations. If you have to > turn off delayed acks to get good nfs performance, then there is a bug in > the tcp stack. See, for example, rfc 2525 page 40, "Stretch ACK violation." > MacOS has this same problem. > "NODELAY" is to turn of the Nagle algorithm, not delayed acks, and Nagle can interact badly with delayed acks on the peer for some workloads/protocols. I'm not sure NFS is one of those cases, but I suspect it probably is.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?34cb7c840607250309t1b20179icb034cad7e720e7f>