Date: Mon, 5 Oct 1998 14:32:56 -0700 (PDT) From: Christopher Nielsen <enkhyl@hayseed.net> To: David Greenman <dg@root.com> Cc: "Mikhail A. Sokolov" <mishania@demos.net>, current@FreeBSD.ORG Subject: Re: -current panics.. Message-ID: <Pine.LNX.4.04.9810051426560.13926-100000@hillbilly.hayseed.net> In-Reply-To: <199810051838.LAA14146@implode.root.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, 5 Oct 1998, David Greenman wrote: > ># Making FreeBSD more dynamically configuring has always been a long term > ># goal. There isn't any magic configuration option that I can make for "big > ># systems", because there are many factors to consider and the system simply > ># has to be tuned to the specific work load. The analysis and tuning is complex > ># enough that it's something that I charge money to do for people. > > > >There was a 2.1.0 when we used to have the same squid with approximately the > >same quantity of clients and there was working dealloc methodic which was gone > >in 2.2.x and, of course, 3.0. It now looks like there's still problems with > >mbuf allocation/leeks and so on (see PR's) and the OS will not work when it > >is supposed to be 'a big system' out of the box. Can I read the answer as > >"FreeBSD is not supposed to be working as huge networking servers out of > >the box"? > > Properly tuned, FreeBSD will work just fine for large servers. You have to > know what you are doing, however. This will be the case for ALL operating > systems, not just FreeBSD. I'm not convinced that there are any "leaks" in > FreeBSD. All of those that I have investigated have turned out to be just > a large average number of TIME_WAIT connections that haven't timed out yet. > The system must be configured with this in mind or you will run out of > buffers. This same kind of situation can occur with a network-intensive app running under Solaris, too. From my own experience with tuning OSs (specifically Solaris 2.x), I'd agree with David's assessment. Not having delved that far into the details of FreeBSD "leaks" as he probably has, I can't say with certainty that there are none, but I'm inclined to believe him. :-) -- Christopher Nielsen Scient: The Art and Science of Electronic Business cnielsen@scient.com <http://www.scient.com>; To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.LNX.4.04.9810051426560.13926-100000>