Hello,
In the literature I found a few ratios aiming to compute the 'offload efficiency'
of your Exadata/SuperCluster setup, but I was wondering if, before beginning
to use various columns in V$SQLAREA it wouldn't be wise in the first place
to check whether, amongst all your SQL statements, how many of them are
candidate to offloading... ?
Therefore thinking to first doing something like
select count(*) into TOTAL
from v$sqlarea where PARSING_SCHEMA_NAME
not like 'SYS%' and
command_type=3 and IO_CELL_OFFLOAD_ELIGIBLE_BYTES=0;
and
select count(*) into CANDID
from v$sqlarea where PARSING_SCHEMA_NAME
not like 'SYS%' and
command_type=3 and IO_CELL_OFFLOAD_ELIGIBLE_BYTES>0;
then compute the ratio of candid / total: in this instance if you end up with a
(very) low proportion, isn't it a sign that in your case Exadata wasn't worth it ?
Or even next to useless if you find ratios kinda like 1.48%, or inferior to 1% ?
What do you think ?
Thanks for any help, hint or englightenment...
Regards,
Seb