Date: Tue, 31 Jul 2001 17:33:15 -0400 (EDT) From: Andrew Gallatin <gallatin@cs.duke.edu> To: Joerg Micheel <joerg@cs.waikato.ac.nz> Cc: hardware@freebsd.org Subject: Re: 3ware Escalade 7xxx supported? Message-ID: <15207.9243.594273.686530@grasshopper.cs.duke.edu> In-Reply-To: <20010801090829.B17381@cs.waikato.ac.nz> References: <3B6597A3.2EDC3A9C@we.lc.ehu.es> <200107310755.f6V7thd01635@mass.dis.org> <20010801090829.B17381@cs.waikato.ac.nz>
next in thread | previous in thread | raw e-mail | index | archive | help
Joerg Micheel writes:
> On Tue, Jul 31, 2001 at 12:55:43AM -0700, Mike Smith wrote:
> > > I did a simple read performance test (dd if=/dev/twed0 of=/dev/null bs=64k)
> > > and I got about 55 MB/s. Hey, that's good :-)
> >
> > Yes. 8)
>
> What do people do if they need 60, 90 or 120 MByte/sec throughput ?
> I have done some experiments with SCSI and there appears to be a
> cap around 80 MByte/sec, regardless how many disks or controllers
> are involved (not a strict FreeBSD problem, only).
>
If you're talking SCSI, get more SCSI buses. Since you said you were
using multiple controllers, you're probably limited by a 32-bit/33MHz
(132MB/sec) PCI bus, so get a faster bus. Eg, get a server with a
64-bit and/or a 66MHz bus.
For example, on a dual-channel sym1010 card in a 64-bit / 33MHz PCI
slot on a Dell PowerEdge 4400, we see roughly 23.5 MB/sec from 8 18 GB
Cheetahs (ST318404LC) split evenly across the buses (or nearly 190MB/s)
<5:22pm>apathy/gallatin:~>iostat -w 1 da0 da1 da2 da3 da4 da5 da6 da7
tty da0 da1 da2 da3 da4 da5 da6 da7 cpu
tin tout KB/t tps MB/s KB/t tps MB/s KB/t tps MB/s KB/t tps MB/s KB/t tps MB/s KB/t tps MB/s KB/t tps MB/s KB/t tps MB/s us ni sy in id
0 2 64.00 1299 81.17 0.00 0 0.00 0.00 0 0.00 0.00 0 0.00 0.00 0 0.00 64.00 1299 81.17 0.00 0 0.00 0.00 0 0.00 0 0 0 0100
0 494 64.00 371 23.22 64.00 372 23.28 64.00 371 23.22 64.00 372 23.28 64.00 368 23.03 64.00 368 23.03 64.00 368 23.03 64.00 368 23.03 1 0 9 3 87
0 422 64.00 382 23.89 64.00 382 23.89 64.00 383 23.95 64.00 370 23.14 64.00 368 23.02 64.00 367 22.96 64.00 367 22.96 64.00 368 23.02 0 0 10 4 86
0 391 63.85 377 23.52 64.00 381 23.82 64.00 381 23.82 64.00 381 23.82 64.00 367 22.96 64.00 367 22.96 64.00 368 23.02 64.00 367 22.96 0 0 12 4 84
0 383 64.00 381 23.83 64.00 381 23.83 64.00 381 23.83 64.00 381 23.83 64.00 374 23.39 64.00 374 23.39 64.00 373 23.33 64.00 374 23.39 0 0 11 4 85
0 402 64.00 380 23.76 64.00 380 23.76 64.00 379 23.70 64.00 380 23.76 64.00 380 23.76 64.00 380 23.76 64.00 380 23.76 64.00 379 23.70 2 0 9 1 88
0 379 64.00 380 23.76 64.00 380 23.76 64.00 380 23.76 64.00 379 23.70 64.00 379 23.70 64.00 380 23.76 64.00 380 23.76 64.00 380 23.76 0 0 9 1 90
0 320 64.00 379 23.70 64.00 379 23.70 64.00 379 23.70 64.00 380 23.76 64.00 380 23.76 64.00 380 23.76 64.00 380 23.76 64.00 380 23.76 1 0 10 0 89
0 426 64.00 375 23.45 64.00 376 23.52 64.00 376 23.52 64.00 375 23.45 64.00 385 24.07 64.00 384 24.01 64.00 385 24.07 64.00 385 24.07 2 0 11 2 86
0 398 64.00 376 23.52 64.00 375 23.45 64.00 375 23.45 64.00 376 23.52 64.00 385 24.07 64.00 385 24.07 64.00 385 24.07 64.00 385 24.07 0 0 11 4 85
^C
This is up from using the on-board aic7899, which is behind a ppb on a
32-bit / 33MHz bus (!!) on the 4400 & we never saw more than 100MB/sec from it.
Drew
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-hardware" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?15207.9243.594273.686530>