Date: Mon, 08 Mar 2004 13:40:20 -0800 From: "Kevin Oberman" <oberman@es.net> To: Steve Kargl <sgk@troutmask.apl.washington.edu> Cc: freebsd-net@freebsd.org Subject: Re: My planned work on networking stack Message-ID: <20040308214020.73C245D07@ptavv.es.net> In-Reply-To: Your message of "Mon, 08 Mar 2004 13:22:55 PST." <20040308212255.GA52526@troutmask.apl.washington.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
> Date: Mon, 8 Mar 2004 13:22:55 -0800 > From: Steve Kargl <sgk@troutmask.apl.washington.edu> > > On Mon, Mar 08, 2004 at 12:22:10PM -0800, Brooks Davis wrote: > > On Mon, Mar 08, 2004 at 10:24:31AM -0800, Kevin Oberman wrote: > > > > > > Unfortunately, SACK is often looked upon as a waste of effort to those > > > who use nets in more commercial forms where aggregation of lots of small > > > streams is how fat pipes are used. Research big science are about the > > > only ones who have a real need for this kind of performance and it's > > > growing fast. Without SACK, FreeBSD will be a non-starter for these > > > purposes. > > > > I've got a co-worker who is part of a research group at ISI that > > is doing research on long fat pipes with large streams. They are > > intrested in doing a SACK implementation. I hope to have some more > > information later this week. > > > > Has anyone looked at Luigi's stuff? > > http://info.iet.unipi.it/~luigi/sack.html > > The page states that Luigi had SACK available in FreeBSD 2.1R, > which was released 8 years ago. I am aware of at least 3 implementations of SACK for FreeBSD over the years, none of which ever made it into the system. At least one of the people who submitted a patch (his was for 2.2) also provided some significant enhancements to one of the GigE drivers that was done with DOE funding to support the LBNL developed BRO IDS, but which were globally beneficial. In both cases, the patches were ignored by those with commit bits and the person who did the work says that he will no longer bother to submit his work to FreeBSD. I was not using FreeBSD at the time that this happened, so I don't know what, if any, objections were raised to the GigE patches, but I have since seen SACK disparaged by others as a waste of time that is not really needed. Obviously they have no interest in >Gbps streams where we have an interest in >20Gbps streams. I'm not trying to stat a flame war here, but it is frustrating and this initiative for a major network code overhaul makes me hope that something will actually happen. It's just that FreeBSD's network stack was once the best around and it's simply not today. Andre's proposal could go a LONG way toward fixing this and I am eagerly looking forward to if! -- R. Kevin Oberman, Network Engineer Energy Sciences Network (ESnet) Ernest O. Lawrence Berkeley National Laboratory (Berkeley Lab) E-mail: oberman@es.net Phone: +1 510 486-8634
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20040308214020.73C245D07>