Re: Re: Interesting idea

On 10/11/10 11:41 AM, jeffsavit wrote:
>>     Yes, but the I/O bandwidth, mmmmm!  *drool*
>
> Actually, I don't think so. My 4 year old home computer does well over 30MB/second on a single commodity-spec disk drive, enough to saturate two ESCON channels of the 3090 time period, and of course server-class products do much more.  Current server-class disk arrays support tens of thousands of I/Os per second (and sometimes far higher).

   Sure, but the IBM iron I've seen have FAR more than two ESCON 
channels.  I don't know about the 3090 series, but if I recall the specs 
of the z9 correctly (I know, decades of difference, but it's the only 
one that sticks in my mind) that machine can have over FOUR HUNDRED 
ESCON channels.  I'm told they're pretty good about keeping them all 
fed, but I don't have metrics here and don't have hands-on experience 
with MF I/O configurations that large.

   How many channels would an "average" 3090-400 have configured?

> I think old truths about I/O performance and particular hardware platforms have stopped being accurate very long ago.

   Some of them, yes.

> On the other hand, what I would like is a good front panel like from a 360/65 or 370/168. Modern computers lack decent blinking lights!

   Ohhhh yes, right there with you on that!

>>     And those vector processors.
>
> I had one of 'em on a 600S - didn't do much good even then - here, Moore's law has moved on without any doubt.

   Most of my applications have been scientific number-crunching.  We 
currently run that stuff on Cray vector machines.

>>     Oh, if I only had the space...
>
> And the solar power array to keep this guy powered up without breaking the piggy bank!

   Ouch, I'd need a MUCH bigger yard!

            -Dave

-- 
Dave McGuire
Port Charlotte, FL