John,

the answer is it depends. Beyond the 'Rules of thumb' posted here, it helps
to pay attention to your data distribution within your set. This usually
means having a tool like HowMessy to provide you with this info (Bradmark
also has some reporting on this).

I have seen cases were it was necessary to play with different capacities to
get one that hashed well. Mainly an issue of how the keys were developed and
therefore how they worked through the hashing.

(Plug alert) If you have Suprtool or Qedit, Howmessy is included and comes
with a nice little help manual that explains the statistics it reports.

While I am tired and this is probably vague, I would encourage you to look a
little further into looking at your key distribution.

Blessings!

Dave
----- Original Message -----
From: "John Penney" <[log in to unmask]>
To: <[log in to unmask]>
Sent: Thursday, March 28, 2002 6:34 PM
Subject: Image D/B capacities


OK, 3000-L or those of you still with me at this hour:

I seem to remember from a '79 D/B course that there was a percentage, 70 or
75% above which you would experience degraded if not molasses-in-February
performance, if you didn't adjust the dataset capacities.

Anybody can quote the source for this quotation?

TIA



John M Penney
Systems Programmer
Production Services
Information Services Department
Pierce County
Tacoma, WA
253-798-6215
253-798-6680 Fax
[log in to unmask]
www.co.pierce.wa.us

The opinions expressed herein reflect those of the sender.  They do not
necessarily
reflect the opinions of Pierce County unless specifically stated otherwise.
* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *

* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *