Return to BSD News archive
Path: euryale.cc.adfa.oz.au!newshost.anu.edu.au!harbinger.cc.monash.edu.au!news.rmit.EDU.AU!news.unimelb.EDU.AU!munnari.OZ.AU!news.ecn.uoknor.edu!news.eng.convex.com!newshost.convex.com!newsgate.duke.edu!news.mathworks.com!newsfeed.internetmci.com!in2.uu.net!zdc!zdc-e!szdc-e!news From: "John S. Dyson" <dyson@indy.celebration.net> Newsgroups: comp.os.linux.networking,comp.unix.bsd.netbsd.misc,comp.unix.bsd.freebsd.misc Subject: Re: TCP latency Date: Fri, 05 Jul 1996 16:53:35 -0500 Organization: AT&T Lines: 44 Message-ID: <31DD8EDF.1A02@indy.celebration.net> References: <4paedl$4bm@engnews2.Eng.Sun.COM> <31D2F0C6.167EB0E7@inuxs.att.com> <4rfkje$am5@linux.cs.Helsinki.FI> <31DC8EBA.41C67EA6@dyson.iquest.net> <4rjrbi$eqc@cynic.portal.ca> Reply-To: dyson@indy.celebration.net NNTP-Posting-Host: n-pc.celebration.net Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Mailer: Mozilla 3.0b4Gold (WinNT; I) Xref: euryale.cc.adfa.oz.au comp.os.linux.networking:44135 comp.unix.bsd.netbsd.misc:3942 comp.unix.bsd.freebsd.misc:22920 Curt Sampson wrote: > > In article <31DC8EBA.41C67EA6@dyson.iquest.net>, > John S. Dyson <toor@dyson.iquest.net> wrote: > > >With many/most web pages being 1-2K, the transfer rate starts to > >overcome the latency, doesn't it? > > No, it doesn't. You have to get beyond about 1-2K that to get past > the slow-start behaviour. If you have a 512 byte MSS, and a 1500 > byte page, it will come back in three segments. The first segment > is sent, and must be acknowledged before the second and third > segments are sent, thus adding another RTT. > If the segments are in the in-memory-buffer, then the latency is buried more. Most network adapters either have multi packet buffers, or DMA directly into memory. If a system is heavily loaded, there might be multiple packets to process at once. > > >I guess FreeBSD-current makes it up by being faster with the fork/execs > >done by simple www servers. (About 1.1msecs on a properly configured > >P5-166.) > > Most modern HTTP servers pre-fork all their processes anyway, I > should think. > Many CGI scripts do lots of fork/execs. Of course, you are right for simple www pages, the servers are ready being already pre-forked. > > >I guess what I am saying is that the results would look more credible > >with a real load, where the networking code would be exercised more. > > This is certainly true. I'd be interested in seeing benchmarks that > test these OSs with several dozen connections per second. > Not only multiple (parallel) incoming connects per sec, but many pre existing connections. (There are some major scalability issues with that also.) John