[KLUG Hardware] Re: IDE RAID

[Bryan J. Smith]

On Fri, 2002-05-24 at 00:35, Mike Williams wrote:
> Can you elaborate on why?  Lemme take a wild guess:  with IDE only one 
> drive can use the channel at a time so you don't gain anything, and there's 
> extra overhead in setting up a transfer on the second drive?

The AT Attachment (ATA) is just a "dumb" bus that allows the integrated
drive electronics (IDE) device to do block transfers with the CPU (PIO)
or memory (DMA).  If two drives try to transfer data simultaneously,
there will be contention and conflict.  This is detrimental if you are
transferring between two drives on the same bus.  I know, I've seen it
first hand (1/5th the performance).

> I figured if I guessed wrong the actual ranter would come forward.  Problem 
> is those things are expensive.  Fastrack is about $70, but a Supertrack 
> will run more like $200.

Hardware-wise, the $20 Promise Ultra = $70 Promise FastTrak.  On some
models, only a pull-down resistor is required to flash the former with
the latter's BIOS.

> I'd much rather have a 4-drive RAID 5 than a 2-drive stripe, though.

Depends.  4-drive RAID-10 is better than 4-drive RAID-5 at a number of
things, although it "costs" 2 drives instead of just 1.

> I'd be glad to.  Reviewing services seems like a fair trade for you 
> educating me on the subject.

I'll send it off-list since I might be signing over the copyright soon. 
;-P

> Latency per drive would be identical, yes.  Don't you have to wait for the 
> data on all drives to come around, though?

Yes, but they are being accessed simultaneously in the case of RAID-0. 
In the case of RAID-1, you can actually balance independent, multiple
reads from different disks -- *IF* your controller does this (as 3Ware
does ;-).

> I'm not suggesting that you have to add latencies, but the bell curve
> should lean to the right some.

I'm missing your point.

> That's what I was referring to below.  I didn't realize it was a vendor 
> specific thing rather than a true standard.

It's a _model-specific_ thing in many cases.

> Really?  Why waste processor cycles on such a thing if the SCSI controller 
> and / or drive firmware shuffle the write order?

That's the beauty of ATA RAID.  By adding an on-board controller that
does this, you get SCSI-like command tagging and queuing while saving
cost on the drives.  This can only be a "good thing" for the OS, because
it only further augments what the OS does.

> Of course, on most machines read performance is much more important
> than write.

Correct.

> Any idea how that happens?  I can't believe that the rotational speed is so 
> uniform that they don't drift.

They do.  And I'm sure spindle synchronization helps.  But from what I
read years ago, it is small enough that the extra cost is not worth the
2-5% improvement for commodity systems.

> I'll believe that movement of the read/write head is, but that's not
> the issue when we're talking about rotational latency.

Identical disks in a RAID volume are organized so their mirrors/stripes
are nearly identically displaced/contiguous.

-- Bryan

-- 
The US government could be 100x more effective, and 1/100th the
Constitutional worry, if it dictated its policy to Microsoft as
THE MAJOR CUSTOMER it is, and not THE REGULATOR it fails to be.
---------------------------------------------------------------
Bryan J. Smith, SmithConcepts, Inc.   mailto:b.j.smith@ieee.org
Engineers and IT Professionals     http://www.SmithConcepts.com