Re: Followup: Sequence Numbers as Primary Keys -- More Information
Date: 1995/09/27
Message-ID: <DFL3E9.BxK_at_lcpd2.SanDiegoCA.ATTGIS.COM>#1/1
Joel Hagans (jhagans_at_telerama.lm.com) writes, in part:
>1. Speed/Efficiency
>IT group did speed tests of handle objects within their C++ program with the handle
>objects mirroring the PK of the database. This was not done to time the speed of
>Oracle, but the speed of the program (I can elaborate more if you wish -- send e-mail).
>The results of doing 6 million compares with sequence numbers was 13 seconds. With
>the most expensive composite key (6 elements -- 4 varchar2, 1 date, 1 number), it
>took 4 minutes 38 seconds. The program is estimated to do 36 million
>comparisons a day.
The ratio of 13 seconds to 278 seconds is an impressive difference. But not an impressive SAVINGS. The 13/278, if I understand your test, is based 6 million compares. Your production day only anticipates 36 million comparisons a day. That sounds like a rather light application to me. The total daily mission would be just six times as long as your test. So that would be 78 vs. 1668 seconds or, a minute, 18 seconds vs 27 minutes, 48 seconds. So the daily saving is about 26 minutes. Round up to a half an hour, if you like. Not much in the grand scheme of things. Is user response time an issue? If so, how many comparisons are there in user query or transaction? What kind of response time do they expect?
>The DBA group claims that they have followed their rules and guidelines in the design
>of the database. If IT doesn't approve, then don't expect any support from them.
Follow 'em. If you create a performance problem, they'll have to bail you out, right? They OWN performance and don't let'em forget it.
>Therefore the question now is ... Is this still worth pursuing?
No.
--
Martin Sandman Not a spokesperson for .___.
AT&T GIS ESD A3 02 01 Teradata, NCR, or AT&T. |_. |
17095 Via del Campo Martin.Sandman_at_SanDiegoCA.ATTGIS.com .__| |_|
San Diego, CA 92127 Phone: 619-485-3468 | |__| |
Received on Wed Sep 27 1995 - 00:00:00 CET