[rescue] 36G SCA drives still looking

[Charles Shannon Hendrix]

Fri, 04 Mar 2005 @ 14:08 -0500, Kevin Loch said:

> The only ide failure I have had over the past several years is 
> apparently corrupted firmware on disk due to some bugs in FreeBSD 5.2
> (or the drivers with that version)  Having firmware on disk is by far 
> the biggest risk with ide drives.

That, and lower quality control.  IDE drives are where they cut corners
the most.

> I have lost several scsi disks during that timeframe with mechanical
> problems.
> 
> I think the mechanical differences/reliability are greatly exaggerated.

Not always.  Some IDE drives really are mechanically inferior.

Seagate seems to make decent IDE drives, and I've heard that some of the
Barracudas are literally identical except for the electronics.

However, other IDE drives are definitely different.  For example, open
some drives up and look at them inside.  I have and found several IDE
drives that were notably inferior: cheaper metal, less precision, fewer
or smaller air filters, thinner PCBs, cheaper parts, etc.

Another thing which bothers me is that with some IDE drives, they
evidently have firmware that cheats, or lies to the computer.

For example, you tell it to turn write cache off, but it doesn't.
Some people say that is why Maxtor and WD win benchmarks over IBM and
Seagate.

On a BSD mailing list some time ago, there was a discussion about IDE
drives lying, and how it would affect things like journaling and RAID
software.

Some people said that IDE drives could not garantee write order, which
is important for both.  I don't believe they are all that way, but it is
worth thinking about.

Personally, I am fairly sure that a couple of the IDE drives I had never
turned write cache off.  I base that on there being no difference in any
benchmarks I ran, while with other drives it *did* have an effect.

> All drives with unique data should be backed up or made redundant
> regardless of technology.

True.

> My problem with SATA is that while they are vastly superior
> to ATA, they still do not completely match the parralel task
> performance of SCSI.

NCQ is supposed to help.

What I don't understand is why the industry wasted so much time IDE.

SCSI was better, and at the same quantities, the price should be about
the same.

I don't see any reason for IDE to have ever existed.

Also, since IDE drives have to have a controller on each drive,
shouldn't they technically cost more than an identical SCSI drive?

Or do the extra capabilities of SCSI mean that their PCB is just as
complex?


-- 
shannon "AT" widomaker.com -- ["Tara is grass, and behold how Troy lieth
low--And even the English, perchance their hour will come!"]