Return to BSD News archive
Path: euryale.cc.adfa.oz.au!newshost.anu.edu.au!harbinger.cc.monash.edu.au!nntp.coast.net!news.kei.com!newsfeed.internetmci.com!zdc!zdc-e!szdc-e!news From: "John S. Dyson" <toor@dyson.iquest.net> Newsgroups: comp.os.linux.networking,comp.unix.bsd.netbsd.misc,comp.unix.bsd.freebsd.misc Subject: Re: TCP latency (lucid discussion :-)). Date: Mon, 15 Jul 1996 18:33:45 -0500 Organization: John S. Dyson's home machine Lines: 58 Message-ID: <31EAD559.41C67EA6@dyson.iquest.net> References: <4paedl$4bm@engnews2.Eng.Sun.COM> <x7687w1dsr.fsf@oberon.di.fc.ul.pt> <4s220u$nmq@symiserver2.symantec.com> <31E53C2B.41C67EA6@inuxs.att.com> <4s6k8o$4o0@fox.ksu.ksu.edu> <31E6FD92.41C67EA6@dyson.iquest.net> <4s8cuq$ljd@bertrand.ccs.carleton.ca> <31E7C0DD.41C67EA6@dyson.iquest.net> <4s8tcn$jsh@fido.asd.sgi.com> <31E80ACA.167EB0E7@dyson.iquest.net> <4see1t$f0l@fido.asd.sgi.com> NNTP-Posting-Host: dyson.iquest.net Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Mailer: Mozilla 3.0b5aGold (X11; I; FreeBSD 2.2-CURRENT i386) Xref: euryale.cc.adfa.oz.au comp.os.linux.networking:45248 comp.unix.bsd.netbsd.misc:4052 comp.unix.bsd.freebsd.misc:23600 Larry McVoy wrote: > > John S. Dyson (toor@dyson.iquest.net) wrote: > : Larry McVoy wrote: > > : > It's kind of fun to contrast this with the Linux crowd: > : > > : > BSD guys: "your benchmark sucks! your numbers are wrong! you mislead the > : > world! you suck! whine!" > : > > : > Linux guys: "Hey Larry, the null syscall benchmark is busted because we can > : > do a null syscall in about 1 usec. You need to change it so > : > it times in nanoseconds". > : > > : I think that was a kind-of cute situation. We decided NOT > : to special case the syscall that Larry uses for the > : null-syscall case. > > As has been pointed out, neither did Linux. So perhaps you might just ask > yourself, how come Linux is doing so much better? > In that area, it is a difference in VFS overhead, that I have determined not to change at this time. There are more interesting things (interesting in the sense of things that impact real-world performance more) to work on. My attitude is that some benchmarks are mostly interesting to measure some micro-level operation that may or may not affect performance on a larger scale. Some benchmarks measure things on a larger scale, but sometimes make it difficult to track down the performance problems. I use your lmbench suite mostly as a micro level suite that does not necessarily measure real world performance, but simply as a guide post to see what is going on. (In a sense, to help in the process of quality control, but not necessarily as a quality measure in its own right.) Since there was a difference in the performance between Linux and FreeBSD, I had evaluated the difference, and found that it did not generally affect the operation of the system enough to warrant modifications at this time. I did muse about "fixing" the system so that the benchmark would work faster, but that would affect the performance in general as much as making the benchmark actually run faster without tricks. Linus, of course, had misunderstood my intentions on the above remark and quickly accused me of something that I was not aware of doing. Of course, the difference in VFS overhead affects small transfers substantially, but again, most people don't do small transfers. Instead, I am looking at the the benefits that are gained by modifications, and the risks associated with causing our code base to be incompatible with the rest of BSD, and decide to lean towards the conservative in this case. If I find more than a few users that are inconvienienced by our/my choice, then I am sure that we will welcome discussion and participation of the appropriate parties to decide upon the appropriate action. So, my answer to your question is: I don't necessarily use the lmbench benchmarks as the definitive measure of quality. However, it is a good tool to assist in quality control. So specifically, I will answer and say that I don't use 'lat_syscall' as a direct measure of quality. John