Date: Fri, 15 Jun 2001 00:49:20 -0700 From: "Ted Mittelstaedt" <tedm@toybox.placo.com> To: "Szilveszter Adam" <sziszi@petra.hos.u-szeged.hu>, <freebsd-advocacy@FreeBSD.ORG> Subject: RE: An interesting read... Message-ID: <001501c0f56f$af7f3ba0$1401a8c0@tedm.placo.com> In-Reply-To: <20010614092753.A13828@petra.hos.u-szeged.hu>
next in thread | previous in thread | raw e-mail | index | archive | help
>-----Original Message----- >From: owner-freebsd-advocacy@FreeBSD.ORG >[mailto:owner-freebsd-advocacy@FreeBSD.ORG]On Behalf Of Szilveszter Adam >Sent: Thursday, June 14, 2001 12:28 AM >To: freebsd-advocacy@FreeBSD.ORG >Subject: Re: An interesting read... > > >The article is available online: > >http://www.sysadminmag.com/current/0107a/0107a.htm > >and, as was already discussed on the Hungarian FreeBSD users list, it >mostly hits upon the usual suspects: > >- threading is not as efficient on FreeBSD as it is on some other OSs. We > know this (esp in SMP situations). >- File system performance lagging because of sync mounts and no softupdates > (the authors tested with default installs) we know this too. >- It supplies a rehash of the controversy of using separate processes in > daemons to handle new tasks versus using threads and using blocking vs > non blocking TCP/IP calls. The author clearly states his preference for > asynchronous threads and non-blocking TCP/IP calls. (He is from the > company that makes the Lyris mailing list server) >- Author states that in his test (Using 4.2) he was unable to make FreeBSD > serve more than 2500 connections simultaneously while testing email > delivery to a test list. > >Contrary to expectations, w2k is not first on the list, Linux comes in as >winner. But FreeBSD is indeed the last. I think that (while you can argue >that his design choices may not be equally suitable for all network >applications) the issues he raises are partly valid. We have known about >them for a long time. 5.0, if it ever hits the streets:-) will hopefully >correct some of these, provided that softupdates will reach the level of >maturity that you can use them without worries on all file systems. > >The tests (as usual for lab tests) were not necessarily based on real-life >situations (eg writing etc 10000 files in a directory) but they still give >some hints, and this is all such benchmarks are good for. > Benchmarks that don't attempt to mimic real-life situations are not generally useful to consumers. Sure, they give hints as to where the weak areas are that need work, but the fact of the matter is that all operating systems have weak areas that show up when they are pushed to their limits. I don't have a problem with such articles in scientific research journals but SysAdmin magazine is not a scientific research journal. It is a popular trade press magazine aimed at the average system administrator of an average network. The article should have been written with a discussion of factors that is important to this audience, not with a discussion of factors that are important to mailing listserver administrators. (of which there are not many since most people use ISP-hosted mailing lists or one of the free servers intended for low-volume listserving) It would have probably been more useful to the audience if the author had included a sidebar that discussed what happened to each of the servers when their power switches were turned off right in the middle of writing out those 1000 files. Ted Mittelstaedt tedm@toybox.placo.com Author of: The FreeBSD Corporate Networker's Guide Book website: http://www.freebsd-corp-net-guide.com To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-advocacy" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?001501c0f56f$af7f3ba0$1401a8c0>