RE: Operations that perform multiblock I/O and cluster factor

Actually, I have seen "db file scattered read" on INDEX RANGE SCAN, and I had a 10046 trace file for that, but was unable to find it (I must have deleted it).

This was a very degenerate case, where the entire table was accessed through single column index (hinted). The trace file showed around 70% "db file scattered read", 20% "db file sequential read" and around 5% "db file parallel read". This was Oracle 9.2.0.6 on Linux.

Don't know what is causing that, though. The column did have a very bad clustering factor.

<"Allen, Brandon">
> tune.htm#15945). There are other multi-block I/Os though, for example
> direct path reads & writes to temp, but those show up under their own
> wait events, not "db file scattered read", and they do not occur on
> behalf of an index range scan. I think your co-worker is wrong and if
> s/he's going to make such claims in contradiction to the documentation -
> the burden of proof is on him so ask him/her to show you a 10046 trace
> file with "db file scattered read" wait events coming from an INDEX
> RANGE SCAN, or some other proof.

-- 
Riku Räsänen
Kantamestarit OY
www.kantamestarit.fi

--
http://www.freelists.org/webpage/oracle-l