On Wed, 27 Oct 1999, Per Larsson wrote:

> I just want to say that it works perfectly on my uV3100/20e.
> Ragge has done some great work, as usual.
> 
> I got about 350k/s with a little test, which seems somewhat slow compared
> to the numbers thrown around here. This is probably because the poor
> machine only has 4MB RAM. I'm getting more memory soon.

It also depends a lot on the drive you're using and the blocksize you're
using to access the drive.  For instance, if you're hitting the raw device
/dev/rsd0c using `dd' with a block size of 64k vs 4k or 1k, you'll see
significant variance in the throughput.  If you're doing the tests by
accessing files on a filesystem, you'll see variance there depending on
the filesystem, the parameters with which the filesystem was created, the
amount of buffering set aside for filesystem access, the size of the file,
and whether or not you've recently accessed it :-) And of course nice new
multi-gigabyte SCSI drives are a lot faster than say a klunky old RZ22
from DEC.

-brian.
--- Brian Chase | bdc@world.std.com | http://world.std.com/~bdc/ -----
              "You are not expected to understand this." 
   -- A comment from the source of UNIX 6th Ed, unix/slp.c, line 438