Re: IDE raid 1


Subject: Re: IDE raid 1
From: Christopher E. Brown (cbrown@woods.net)
Date: Sat Mar 15 2003 - 12:42:13 AKST


On 14 Mar 2003, Damien Hull wrote:

>
> Looks like I'll have to do more research on RAID cards. The only one I
> was aware of was the Promise card.
>
> It will be a while before I get to the hardware stage. Right now I'm
> still in the planning and testing stage.

Beware of IDE hardware RAID cards, they are *NOT* hardware raid.
Cards such as the promise are a standard UDMA controller with some
extra BIOS. This BIOS can layout the array/etc but does nothing else.
The actual RAID is done within the OS driver for the card.

You are better off getting the non raid version of the card, and using
Linux's native RAID code, instead of the card drivers. The tools
avail for native linux raid are much more flexible.

Also, most motherboard ATA RAID *cannot* be run as non RAID (the
stupid RAID bios insists on setting up ATA raid tables on the drive),
and many ATA raid cards are like this as well.

If you *really* want a cheap RAID, get a UDMA-133 controller supported
by Linux and use Linux native RAID.

Other options are an IDE<->SCSI raid box (like an SX4000), the system
sees it as one big SCSI drive, or IDE<->SCSI converter modules (per
drive, the system sees a pile O SCSI).

I of course prefer a large array of cheetas, but if I need a cheap but
still nice setup I use the per drive converters. The system is all
SCSI (normally U160LVD) out to the converters, but the drives are
fairly cheap IDE. All the benefits of cheap drives, but without the
IDE overhead (IDE performance is way up these days, but there are
still many issues with overhead. In many cases the CPU is left
spinning, waiting on IDE task completion due to the lack of a full
connect/disconnect and command queuing system)

Oh, and IIRC there are some issues in 2.4/2.5 with the Promice PDC
controllers (RAID or straight IDE driver). These are being beat on,
but there are still issue reports happening.
 --
I route, therefore you are.

---------
To unsubscribe, send email to <aklug-request@aklug.org>
with 'unsubscribe' in the message body.



This archive was generated by hypermail 2a23 : Sat Mar 15 2003 - 12:41:38 AKST