RE: maxthr system statistic
- From: "Jonathan Lewis" <jonathan@xxxxxxxxxxxxxxxxxx>
- To: <oracle-l@xxxxxxxxxxxxx>
- Date: Sat, 20 Jan 2007 07:51:27 -0000
Brandon,
I don't know much about this, but for what it's worth,
the MAXTHR goes with the SLAVETHR and is used
is the calculation of cost of parallel execution.
As it says, perhaps slightly misleadingly, in the 10053
trace file in 10gR2
MAXTHR - maximum I/O system throughput
SLAVETHR - average slave I/O throughput
I think this means Oracle is monitoring the rate
at which a query co-ordinator can acquire
data from it's slaves checking for the maximum
(hence the need that Christian saw to reset the
value), and the rates at which the slaves are passing
data to the QC and each other.
I believe that Oracle then uses these values in some
way to adjust the cost of parallel queries by throwing
in a fudge factor that chokes the degree used if it
looks "optimistic" compared to
maxthr /slavethr.
But I don't haven't worked out any details.
Regards
Jonathan Lewis
http://jonathanlewis.wordpress.com
Author: Cost Based Oracle: Fundamentals
http://www.jlcomp.demon.co.uk/cbo_book/ind_book.html
The Co-operative Oracle Users' FAQ
http://www.jlcomp.demon.co.uk/faq/ind_faq.html
----- Original Message -----
Subject: RE: maxthr system statistic
Date: Fri, 19 Jan 2007 11:00:36 -0700
From: "Allen, Brandon" <Brandon.Allen@xxxxxxxxxxx>
Thanks Christian & Mladen for sharing your thoughts, unfortunately they
only confirm my thoughts that nobody really knows how this statistic is
gathered or used - and based on the large variation in values on the
same system, maybe Oracle doesn't even know. I wonder if maybe it's not
used at all currently, but just there for use in a future version where
perhaps it will be better documented? I guess for now the safest
approach is that which you described Chris - gathering stats several
times and then taking the average (I'm guessing that's what you meant by
"good"?) value.
Regards,
Brandon
--
//www.freelists.org/webpage/oracle-l
Other related posts:
