Date: Fri, 29 Sep 2006 08:52:58 -0700 From: Chris <snagit@cbpratt.prohosting.com> To: FreeBSD Questions <freebsd-questions@freebsd.org> Subject: Swap Size Importance? Message-ID: <BAB82621-2B6E-4D53-9B0B-C7158E945E95@cbpratt.prohosting.com>
next in thread | raw e-mail | index | archive | help
As a standard practice, I've always configured swap file to be double the size of real ram split across system and data disk. For example, 8gb on da0 and 8gb on da1 if the system has 8g real ram. In practice, In 7 or 8 years, I've never seen swap used for anything but a few k of inactive processes and I would imagine if real active process swapping occurred, it would be an immediate indicator that the system that isn't responsive enough for use anymore and requires upgrade or tuning. Can't run a website process off disk and keep anyone coming to the site ;-). (BTW, I'm talking only about high end servers, not test boxes where I've seen lots of swapping). I'm at the point of attempting my first gvinum software raid-5 and realized, I need the entire disk storage of all three non-system drives to avoid pulling an 8gb chunk out of the drive sizes. The configuration is one scsi 72g system disk and 3 that will be used for the raid volume. I should mention I turn off dumps, haven't found the use for that in a production server since it should not be rebooting or it's back in the shop and another box is taking it's place. Is there any shortfall in performance or reliability to running production with swap equal in size to the 8gb of system memory? I can't think of any but don't want to make a hard to correct mistake once this thing goes in.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?BAB82621-2B6E-4D53-9B0B-C7158E945E95>