Re: Multi-Block read count

Connor McDonald doodled thusly:

>> >
>> > it is platform dependent. The method I use in order to determine the
>> > maximum in a database is simple, using sqlplus:
>> > * alter session set db_file_multiblock_read_count = 10000;
>> > * show parameters db_file_multiblock_read_count

this may be accurate in some cases but it assumes that ORACLE and the OS have been tuned to fully exploit the max IO size. that is not necessarily always true, particularly the combination of both products.  

>
>Its also worth nothing that the ceiling on multiblock read count is what
>*Oracle* thinks that it will be able to achieve. Its quite possible
>that you will not be able to get that due to other restrictions...

exactly.

>
>For example, Oracle on Solaris can get up to 1m, but unless you set
>maxphys kernel parameter you won't get near that.

precisely. but you'll still see the "right" value in the parameter list, if you use the method above of setting dfmrc=10K.

>
>To see what you can actually get, set a 10046 trace at level 8 and run a
>full scan on a big table. The p3 values in the trace file for the
>scattered reads shows you what Oracle could achieve in reality.
>

yup, that's the way to go to find out exactly what IO size can be achieved.

And even with this I've had some surprise diffs between raw and cooked file access...

Cheers
Nuno Souto
nsouto_at_optushome.com.au.nospam Received on Thu Jan 10 2002 - 03:18:31 CST