Re: DASD performance curve graphs??

Al,

The more arms the better has been a mantra I've subscribed to for my entire
career.  I've noticed a 20% reduction in run time for some jobs once we went 
from
3 drives to 4 and gave the system a little time to load balance.

I might want to throw in that a prudent use of ASP will help your overall
throughput depending on the design and mix of the applications being processed.

In our case we have master/transaction files in one ASP because that is where
most of the interactive activity occurs.  Here the 40% activity threshold is
important.

For the really huge historical files that some analytical process is being done
against, in batch mode, they get put into their own ASP.  Here the 70% space
threshold is important.

If the files were all in the same ASP, even at a lower priority, the nature of
the analytical jobs are large disk I/O and they read until they time out and 
that
can put a real crimp on interactive jobs that must do 20+ small gets per screen
change.

You are right about the cost of manufacturing.  It doesn't cost twice as much to
build a drive with twice the capacity.  Now if they can just get the arms to 
move
a little faster ...

Oh, I just had a geezer pop up, didn't IBM have a drive at one time that had
stationary read/write heads, just a whole lot of them?  Instead of the head
moving from cylinder to cylinder, a different head was activated?  Maybe it's
time for something old to become new again.



"Al Barsa, Jr." wrote:

>
> I misunderstood your comment.
>
> 40% is the percentage that the arms are busy, not the percentage full.
> Look at the column on the right of WRKDSKSTS.  As the disks get full to
> typically 70% or 80%, this will increase the % busy.  The big issue today
> is that if you buy less arms with higher capacity, the percentage busy goes
> up at a very fast rate.  The DASD industry has just figured out that they
> can create a 35GB device in a 3 1/2 form factor for the same cost of a 2 GB
> in the same form factor.
>
> Al

+---
| This is the Midrange System Mailing List!
| To submit a new message, send your mail to MIDRANGE-L@midrange.com.
| To subscribe to this list send email to MIDRANGE-L-SUB@midrange.com.
| To unsubscribe from this list send email to MIDRANGE-L-UNSUB@midrange.com.
| Questions should be directed to the list owner/operator: david@midrange.com
+---