[rescue] Apple Announces Rackmount Server - WITH IDE HARDWARE RAID

[James Lockwood]

On Wed, 15 May 2002, David Passmore wrote:

> On Wed, May 15, 2002 at 02:23:57PM -0400, Joshua D Boyd wrote:
>
> > Of course, do they even meet their own specs?  My IDE drives run nowhere
> > close to the 33 or 66 megs they advertise.  I'd be shocked if ATA100 drives
> > run anywhere near 66mb/s either, but each hard drive has to crank 66mb/s
> > sustained to meet the Apple spec.
>
> It takes several drives to saturate the bandwidth of any kind of bus. The
> one drive per channel thing isn't about bandwidth, it's about contention.
> Similar to how ethernet switching was created to eliminate collisions, not
> increase bandwidth.

Arguably more important than that is the fact that IDE hotswap would be
very difficult to implement properly with both a master and slave drive on
one chain.  For that matter, graceful degradation would be tough.

My beef with Apple over using IDE is largely mitigated by their "one
controller per drive" design decision.

-James