You'd have to do some really fine monitoring and tuning to identify that the
performance impact( better or worse) is strictly the impact of changing
these values.
They're used by the optimizer to identify the I/O expected workload so that
it would choose an access plan that would not be optimal because the parms.
were not set right.
The only way I think to test this out is to ALTER your tblspcs. values. Take
an explain before and after and see if it makes a change.
If both call for a relational scan, then prefetching will happen and it will
happen at the true rate of transfer at execution time so a benchmark would
probably be the only way to find out.
You'd have to snapshot before and after the change, collect the I/O times as
well as number of pages read and see if there's realy an impact.
This also supposes that you've tuned everything else as best as possible
(num-ioserver, b uffer pools and so on...)
HTH, Pierre.
--
Pierre Saint-Jacques
SES Consultants Inc.
514-737-4515
"Jean-Marc Blaise" <no****@nowhere.com> a écrit dans le message de news:
42**********************@news.wanadoo.fr...
Hi Pierre,
Well, even if not everybody uses ESS, this kind of storage is not
obsolete,
and I am surprised that IBM recommandations for ESS that you can find in
2001 & 2003 redbooks, so not so old, are completely dephased with the new
defaults.
Speaking of "an acceptable average value for current hardware", I would
have
imagined a higher value like 0.4 or 0.5 (the average for ESS independantly
from page size gives 0.625). I wonder the impact on optimization if the
default values tell that your disk is greater than it really is.
Regards,
Jean-Marc
"Pierre Saint-Jacques" <se*****@invalid.net> a écrit dans le message de
news:EV********************@weber.videotron.net... If I remember correctly, the default values were introduced in DB2 Common
Server V2.0 (1993-4?).
They were based on current (then) hardware technology.
I think the technology has evolced!
To your No. 1) Not everybody uses ESS, and I think the new defaults
reflect an acceptable average value for current hardware.
To your No. 2) I thought so but now I'm not sure...
HTH, Pierre.
--
Pierre Saint-Jacques
SES Consultants Inc.
514-737-4515
"Jean-Marc Blaise" <no****@nowhere.com> a écrit dans le message de news:
42**********************@news.wanadoo.fr... > Hi folks,
>
> Transferrate & Overhead's defaults have been reduced in version 8.2,
> respectively 0.9 to 0.18 and 24.1 to 12.67.
>
> 1) Why reducing so much the transferrate so that it seems ridiculous
> compared to values preconised by IBM for ESS storage :
> 4K -> 0.1
> 8K -> 0.3
> 16K -> 0.7
> 32K -> 1.4
>
> 2) Why is the transferrate default independant of the page size ?
>
> Thanks for your help,
>
> Jean-Marc
>
>
