Re: RAID information

In article <1993Sep13.042159.8855_at_a.cs.okstate.edu>, minich_at_a.cs.okstate.edu (Robert J Minich) writes:
>Newsgroups: comp.arch.storage,comp.databases.oracle,comp.sys.hp,comp.sys.dec
>From: minich_at_a.cs.okstate.edu (Robert J Minich)
>Subject: Re: RAID information
>
>by RAID7_at_world.std.com (John OBrien):
>> I've Emailed out our detailed technical studies explaining the different

                    ^^^^^^^^ ^^^^^^^^^


>> RAID levels to those people requesting copies.  Sorry for the delay.  I am


>> not in the office so much nowadays.  I expect to have another technical

>> report published in the literature shortly.

And here, I've always thought it was a RAID-4. The distinction between the two depends on whether the data is bit/byte interleaved across the devices or block interleaved.

It has been the past claim of Mr. OBrien that the distinguishing features of the different Berkeley RAID levels, were based on the performance differences and not the topological and reliability differences. Those disputing Mr. OBrien's claim are likely to say that the performance differences are *due* to the topological differences.

The computer storage architects can argue the details of the point for years and probably never resolve it. That might change of course if an author of the original papers were to look at RAID-7 and say "Looks like RAID-4 with lots of cache thrown at the hard parts to me...". Such arguments while entertaining, are of little value to customer (*).

What really matters is whether Mr. OBrien's product offers the desired performance, data reliability and data availablility at an acceptable price.

(*) The classic exception is that if most of the rest of the world concludes it is a RAID-3 or RAID-4 with lots of cache thrown at slow parts, it hurts Mr. OBrien's credibility, which in turn may hurt his product.

>--
>Robert Minich
>minich_at_a.cs.okstate.edu | "Maybe awake is what's left at the end
>Oklahoma State University | of dreaming"
>

--
Alan Rollow				alan_at_nabeth.cxo.dec.com