[geeks] was Re: [rescue] Small RAID array setup

[Andrew Weiss]

On Friday, June 27, 2003, at 01:21 PM, Phil Stracchino wrote:
>
>
> The mechanical complexity boggles the mind, and I see no advantage that
> it offers over a single spindle.  Have you considered:
>
> (1) You now need two separate spindle motors?

Yes...there would be two of everything and they talk through a shared 
arch.  The two spindle poles are close together because the platters 
alternate so they are almost 1.5 x disc diameter in drive length at 
least... so longer drives perhaps.

> (2) You gain nothing in access speed unless you have two controller
>     boards driving two separate steppers?

Yes...

> (3) Disk heads "fly" on the boundary-layer airstream, and need a clean
>     boundary layer do to it.  Have you considered the turbulence 
> greated
>     by two platters spinning in opposite directions at 10000rpm a few
>     millimeters apart?

No

> (4) Internal mirroring like this would be utterly pointless.  If one
>     half of the drive loses power, so does the other.  If one half has 
> a
>     head crash, it's GOING to take the other half with it, guaranteed.

Not really...a lazy cache takes care of the power issue....also a hot 
plug controller board with drive geometries allows upgrades.  Lastly .. 
I didn't say there wouldn't be two power connectors... If one set of 
heads crashes it doesn't affect the second head set for the other 
spindles... they are on opposite sides of the drive.

(5) This scheme would be less reliable than a single disk, because
>     there's twice as many moving mechanical assemblies to fail (at a
>     first approximation), but you still have to swap out the entire 
> disk
>     if one half fails.

No you don't... that was the point... it's a mini hardware mirror in 
the SCOPE of a slightly larger spindle box.

The point of this is you could use it in hardware RAID like a normal 
disk drive, but those who can't afford more than one disk drive or RAID 
options... don't lose their data all the time... i.e. IDE single disk 
PC's.

> (6) Increased size, increased power dissipation, increased heat,
>     increased noise, increased cost ....
>
> I could go on, but I think that's quite enough.
>
Yes, Yes, Yes, Yes, and Yes... nobody said making better things should 
decrease cost.... besides this was just an idea.... if every increase 
was acceptable it might be a good thing...

Andrew