Date: Thu, 12 Jan 1995 16:06:20 -0500 (EST) From: James Robinson <james@hermes.cybernetics.net> To: steveg@wavefront.wti.com (Steve Galle) Cc: FreeBSD-Questions@FreeBSD.org Subject: Re: mfs file system speed Message-ID: <199501122106.QAA04387@hermes.cybernetics.net> In-Reply-To: <9501120957.ZM2720@walrus.wti.com> from "Steve Galle" at Jan 12, 95 09:57:08 am
next in thread | previous in thread | raw e-mail | index | archive | help
I am sure that if I am wrong that someone will correct me :-), but... Since mfs is ultimately backed by swap space, its speed is found by using RAM. Since you are writing such a big file (something like 25M?), you are forcing your machine to swap. MFS is best used for storing temporary files such as the ones that GCC spits out durring a compile -- if you have enough free RAM, then those files never have to hit disk, hence compiles happen much quicker. I've seen one problem with MFS, though. Say I want to print a really big 3 -> 4M postscript file. If I lpd it directly, it tends to error out saying that the file is too large (is this tuneable? I've not RT(right)FM on that one). So, I use lpr -s, which makes a sym link to the file as opposed to copying the file to the print spool area. When I try that trick with something that is in my /tmp (a MFS), I get the following error: "Your printe job was not printed because it was not linked to the original file." Oh well! James
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199501122106.QAA04387>