Re: Direct load mystery
From: John Hurley <hurleyjohnb_at_yahoo.com>
Date: Wed, 1 Sep 2010 12:56:28 -0700 (PDT)
Message-ID: <2a8d123c-ce75-4505-9d19-3841af9916e2_at_i13g2000yqd.googlegroups.com>
Mladen:
Date: Wed, 1 Sep 2010 12:56:28 -0700 (PDT)
Message-ID: <2a8d123c-ce75-4505-9d19-3841af9916e2_at_i13g2000yqd.googlegroups.com>
Mladen:
> A colleague of mine contacted me with the following issue. She's loading
> data, large quantities of it, into a 64-bit Oracle 10.2.0.5 on SUN
> Slowaris (should I say Oracle Slowaris?). She noticed that there is a
> significant increase in performance if the transit tables she's loading
> the data into are analyzed after they're truncated, immediately before
> the load.
At my shop we delete table stats with cascade set after every truncate table.
The truncate table leaves around the old stats.
I have also seen issues from time to time with tables that were analyzed empty and at my shop we are using dynamic sampling for truncated tables until at some future point actual data is out there. Received on Wed Sep 01 2010 - 14:56:28 CDT