[svlug] Re: 3ware IDE RAID Recommendation ...

Bryan -TheBS- Smith thebs at theseus.com
Tue May 30 13:47:33 PDT 2000

On Tue, 30 May 2000, Rick Moen wrote:
> Sorry, I don't care if they're coprocessed:  Using one IRQ per drive
> pair is not OK.  Neither is lack of hot-fix on the drives.  Nor:

Huh?  Are you familiar with 3ware's products???  These are NOT more
Promise (among other vendors) ATAPI/IDE controllers with some
firmware/BIOS.  One IRQ per board, whether it is 2-channel or

> Sorry, but if you're using ATA disks, by definition you cannot properly
> implement SCSI-style command optimisation.

Not on the disks, but on the adapter itself, YOU CAN!  The
controller looks and acts like a SCSI card.

> Elevator seeking is not possible, for example.

Of course not.  But understand some ATAPI/IDE controllers *DO* have
some SCSI-like features such as sector remapping, etc...  Again,
the on-board processor does quite a bit.

> The host adapter lacks the ability to properly manage pending
> requests, because of the relatively primative adapter-to-drive
> bus.

This is where the on-board processor comes in (it is NOT just the
controllers for the channels).  Unlike other ATAPI controllers,
it does the block transfers to/from the drives.  Then the OS talks
to the co-processor is a SCSI host adapter.  It's NOT a "dumb"
controller card.

> Efficient of money:  As I said, maximum possible redundant storage per
> buck, at the expense of every other design objective.  


I would use this solution *OVER* a bunch of [SCSI] disks (BODs).

> I doubt that, extremely. 

Er, over SCSI disks of the same spindle and density?!?!?!  It is
upto 80% faster!!!

> >>TheBS>> ...when you can't afford SCSI RAID.
> Save up.

Whatever.  I think you're missing my point.  It's not a White/Black
argument here.

-- TheBS

 Bryan "TheBS" Smith -- Engineer, IT Professional and Hacker
      E-mail:  mailto:thebs at theseus.com,b.j.smith at ieee.org
  Disclaimer:  http://www.SmithConcepts.com/legal.html
  TheBS ... Serving E-mail filters to /dev/null since 1989

More information about the svlug mailing list