Jeff Waugh wrote:

2) Plenty of memory negates most of a filesystems performance concerns.

Whenever you're doing fs benchmarks, work with 2 * RAM worth of data, otherwise you're just benchmarking the cache. ;-)

Yes, Jeff.


The corrolary to which is (what he said):  If I have X amount of
data and X amount of RAM, and you have Y*512 amount of data and Y
amount of RAM, then my filesystem/OS/disk/kernel is better than
your filesystem/OS/disk/kernel.

Or, to put it another way:  Mine's faster than yours because mine's
a red one.

I expect that many (insert name of corporation here) funded benchmarks
make very good use of the amount of memory available to disk caches
on various systems.

People involved in *serious* network benchmarking, however, first
make sure that the cables on their test systems are all the same
length.  Then they work up from there.

--
Del

--
SLUG - Sydney Linux User's Group - http://slug.org.au/
More Info: http://lists.slug.org.au/listinfo/slug

Reply via email to