Robert G. Brown
rgb at phy.duke.edu
Thu Feb 28 22:29:16 PST 2002
On Thu, 28 Feb 2002, Andrew Klaassen wrote:
> On Thu, Feb 28, 2002 at 04:07:25PM -0500,
> Robert G. Brown wrote:
> > We just got a Promise UltraTrak100 TX8 (IDE-SCSI) RAID chassis
> > that advertised decent itself as OS-independent plug and play
> > -- attach to SCSI bus and go. The first unit we were shipped
> > didn't work under any OS. The second we were shipped we got
> > the vendor (Megahaus) to verify function before shipping and
> > it does "work", but it returns unbelieveably poor performance
> > at RAID 5 -- a (very) few MB/sec -- under bonnie.
> I think the Promise rep quoted us 12MB/s under RAID 5 for the
> UltraTrak100 TX8. We didn't bite.
In raid 0, under WinNT only. Maybe. We're getting more like 3-5
MB/sec. I suppose it is all my fault, really -- I didn't notice that
the Promise technical docs were the only ones I looked at that mentioned
all the glitter about their system BUT the speed.
I honestly don't see how they could engineer something with performance
that poor without trying.
> > IDE is cheap and easy, but we would like a bit of assurance
> > that linux won't have (e.g. DMA) problems when dealing with
> > 6-8 ide controllers on one bus. Is anyone doing this?
> Lots of people. There's even a mailing list for it:
> linux-ide-arrays-subscribe at lists.math.uh.edu
> ...though I don't know if the archives are online anywhere yet.
> Specific things to know, off the top of my head:
> 75GB IBM drives from the 75GXP line have given a lot of people a
> lot of problems. (Apparently it's a firmware thing with earlier
> drives, or something like that. At this company, we traded all
> of ours in for 80GB Seagates.)
> Two Promise Ultra100 cards (not the hardware raid cards; I'm not
> really familiar with any of those) seems to be the max in one
> system. We've used onboard controllers and SIIG cards alongside
> the Promise cards to get the additional ports we've needed.
> Stick with 18" cables. Longer cables work for some people, but
> I've never had success with them. Short cables make it a little
> harder to find a suitable case, but these cases:
> ...will keep up to five drives properly cooled, in easy reach of
> IDE controllers. With a little imagination and an extra fan or
> two, you could easily get 8 or 12 well-cooled drives in a case.
> (We've currently got 8 drives in one of these cases in
> production. We had to drill extra holes in the center cage to
> get proper spacing between the drives, and mount an extra fan in
> In the last few days, people have started talking about problems
> with 3ware cards (which still seem to be, despite that, the
> best-supported hardware RAID IDE controllers available).
> Unfortunately, I haven't really been following that discussion,
> so I can't give any more details.
An immensely helpful reply, thanks. Most repliers praised the 3ware
controller, in some cases lavishly so we're going to look into this
solution first. Too bad the list isn't mailman based -- then there
would be an archive -- but if we don't come to a resolution right away
on the basis of what we've learned in the next day or so I'll (sigh:-)
add it to the eight or so other lists I'm already on.
Robert G. Brown http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567 Fax: 919-660-2525 email:rgb at phy.duke.edu
More information about the Beowulf