Re: Q: SQL*Loader Performance

From: Steve Butler <sbut-is_at_seatimes.com>
Date: Fri, 24 Mar 1995 08:34:11 -0800
Message-ID: <Pine.SUN.3.91.950324082628.25719A-100000_at_seatimes>


On Mon, 20 Mar 1995, Murray Kaiser wrote:
> With both databases up and running, I get the following statistics:
> Elapsed time was : 00:29:44.19
> CPU time was: 00:16:26.82
>
> With only one database up and running, my statistics are:
> Elapsed time was: 00:41:00.58
> CPU time was: 00:38:34.95

  1. You were loading the SAME database each time?
  2. Physical files were at the EXACT same physical location on each drive?
  3. Was one load using DIRECT path and the other didn't?

I've seen examples of file placement causing a doubling of run time. It was traced down to the rotational delay on reading data off the drive. For the fast run, everything was positioned such that the data was almost ready to flow past the read head when needed. For the slow run, it has just gone past the read head and the system had to wait for the rotational delay of the disk.

These are the types of things that a good pre-fetch caching system are supposed to fix. Perhaps one of the other users was managing to get the blocks your load needed into memory before your loader needed them. Thats a bit farfetched, but....
--Steve

+----------------------------------------------------+
| Steve Butler          Voice:  206-464-2998         |
| The Seattle Times       Fax:  206-382-8898         |
| PO Box 70          Internet:  sbut-is_at_seatimes.com |
| Seattle, WA 98111    Packet:  KG7JE_at_N6EQZ.WA       |
+----------------------------------------------------+
All standard and non-standard disclaimers apply. All other sources are annonymous. Received on Fri Mar 24 1995 - 17:34:11 CET

Original text of this message