All of lore.kernel.org
 help / color / mirror / Atom feed
From: Ming Zhang <mingz@ele.uri.edu>
To: KELEMEN Peter <Peter.Kelemen@cern.ch>
Cc: linux-raid@vger.kernel.org
Subject: Re: 3Ware 7506-8, anyone have any luck?  I don't...
Date: Fri, 15 Oct 2004 15:19:18 -0400	[thread overview]
Message-ID: <1097867958.3176.97.camel@localhost.localdomain> (raw)
In-Reply-To: <20041015184410.GC3035@inara.maison.net>

Can you explain a bit that with what kind of configuration you can get
that 770MiB/s performance? Thanks.

ming

On Fri, 2004-10-15 at 14:44, KELEMEN Peter wrote:
> [ Please do not address answers directly to me; I read the list. ]
> 
> * Scott T. Smith (scott@gelatinous.com) [20041015 11:21]:
> 
> > You know what, I think I was wrong -- we have the 8506.  Sorry,
> > my bad.  And ours is SATA, not PATA (is that the difference
> > between the 7xxx and 8xxx?)
> 
> Yes.
> 
> > There are other reasons the Highpoint is better than the 3ware,
> > including the fact that it instantly recognizes when you yank a
> > disk, unlike the 3ware which sits there for a while, and then
> > hard resets the entire controller (thus halting all disk io to
> > that controller for a couple of seconds).  3ware wrote me back
> > and told me it's supposed to do that!
> 
> Cannot comment on this one, never was a problem for us.
> 
> > Is the 9500 a PCI-X card?
> 
> No, it's 64-bit 66 MHz PCI.
> http://www.3ware.com/products/pdf/9000DS_041904.pdf
> 
> > [...] and we get the same crappy performance from that one too.
> 
> Define `crappy': I get 770 MiB/s read with 3x 3ware 9500S-8MI.
> (RAID00 hw-sw)
> 
> > Worse yet, the 9xxx series wants to hijack your disks unless you
> > go out of your way to enable 'export JBOD' mode.  You can't take
> > a disk from a 9xxx controller and put it in a box with an 8506
> > -- the controller won't recognize it!
> 
> Have you tried zapping the first and the last megabyte of the
> disk?
> 
> > BTW I'm not talking about RAID or filesystem performance, I'm
> > talking about direct disk access, JBOD mode, using rather large
> > blocks.
> 
> Direct disk access gives me nominal disk throughput on 3ware.
> 
> > So really all I want is multiple controllers on a single card.
> > To that end, the 1820a works great.  I figure we'd all be
> > running software RAID anyways ;-)
> 
> As I said, attractive pricing.  Cannot get stellar hardware RAID
> performance with crappy chipsets. :-)  There's a turnover point
> where you are killed by PCI bandwith limit when using a lot of
> disks.
> 
> Peter


  parent reply	other threads:[~2004-10-15 19:19 UTC|newest]

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2004-10-15 13:23 Slow software RAID5 linux-raid2eran
2004-10-15 17:27 ` 3Ware 7506-8, anyone have any luck? I don't buggz
2004-10-15 17:40   ` Scott T. Smith
2004-10-15 18:12     ` KELEMEN Peter
2004-10-15 18:21       ` Scott T. Smith
2004-10-15 18:44         ` KELEMEN Peter
2004-10-15 18:55           ` Scott T. Smith
2004-10-15 19:28             ` KELEMEN Peter
2004-10-15 19:19           ` Ming Zhang [this message]
2004-10-16 21:37             ` KELEMEN Peter

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1097867958.3176.97.camel@localhost.localdomain \
    --to=mingz@ele.uri.edu \
    --cc=Peter.Kelemen@cern.ch \
    --cc=linux-raid@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.