On Sun, 29 Oct 2000, Stephen Donnelly wrote:
> Alan Cox wrote:
> >
> > > modern IDE disks can only do about 16-20 MB/s *sustained writes*. At the
> > > hub, I've seen a 43GB SCSI disk do 22MB/s, and tailing off after that. A
> > > 10,000rpm 9GB scsi disk was performing similarly I seem to recall. If
> >
> > Remember that as density goes up so does speed when not seeking.. 9Gb disks
> > are much slower as they get less bits/rotation.
>
> Right, I was comparing 43 GB 5400rpm disks to a 9GB 10000rpm drive, they
> preform about the same due to the higher linear density on the larger
> disks, as you said.
If you want a more modern comparison, these are bonnie bechmarks for a dual
PII-400 (or PIII-550, it's been a while) with two 18GB 10k RPM IBM 18LZX
drives. At the time, they were IBM's fastest SCSI drives, but they have
faster ones now. The first is for 1 drive, the second is for two in a RAID-0
config with the Linux md 0.90 driver. The test file was something like 1GB or
so.
-------Sequential Output-------- ---Sequential Input-- --Random--
-Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU
1024 10364 96.8 27659 25.2 12699 24.2 10748 92.3 27221 13.8 369.3 2.5
-------Sequential Output-------- ---Sequential Input-- --Random--
-Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU
1024 10691 99.6 60262 55.4 20568 42.3 11070 95.5 50852 31.5 416.7 3.4
Clearly, sustained output of 25MB/sec or more is possible. Of course, if you
want some spare CPU and PCI/memory bandwidth, that's another thing.
