[linux-elitists] Performance issues with swap file vs. swap partition

Gary Sandine linux-elitists@laclinux.com
Thu Nov 6 11:44:45 PST 2003

On Thu, 2003-11-06 at 12:22, Matthew Galgoci wrote:
> I've actually wondered about this before, and as far as
> I've been told that the physical location of start and
> end of the drive are undefined and not necessarily
> where you might intuitively expect them to be.

I used to do this kind of test a lot when considering partition schemes,
and using bonnie++ 4xRAM I consistently saw stuff like:

"end" of drive per fdisk -- partition at end of drive:
 read: 29414 K/sec
write: 17887 K/sec

"front" of drive per fdisk -- partition at front of drive:
 read: 49934 K/sec
write: 28895 K/sec

IBM 120G ATA 100 drive on an Intel ICH4, done ~8-9 months ago.

The difference is larger in this example than in many similar tests we
logged, but the consistency led me to guess that "front" of drive per
fdisk tends towards outer cylinders.

More information about the linux-elitists mailing list