> Maybe you could give us some more details. How did you do the runstats on
> the fed server? What's your table/nickname, indexes (on both ends), the
> query and the access plan for it? How did you determine that all rows are
> to be fetched from the data source to the fed server?
>
> Which version of DB2 are you using on which platform? Is your fed server
> and remote data source both a DB2 system?
>
> --
> Knut Stolze
> DB2 z/OS Utilities Development
> IBM Germany

Hi
Both servers are runnning db2, but the federator is on 8.2 while the
remote source is in 9.4. I ran a sql from my test db(not the source)
and did a db2explain, it showed me that the predicate evaluation was
done in remote server. But the way our developers were doing was to
construct this sql on the fly, open a cursor inside a stored
procedure , like for example, their input data matches some rules,
they give one particular sql for constructing a cursor with. If it
uses the remote data source (only that with some predicates) , the
target tries to get entire table and do the predicate evaluation.
I have constructed indexes on this nickname (definition only) , that
didnt help me either. I am running out of options.
The way I find that predicate evaluation is done in target is the
cost of query that runs in my source (145MIL -aah..) and the sql from
the snapshot. The sql doesnt have any predicates, and this escalates
to a table level lock as soon as it starts.
Any help would be appreciated.
Thank you
Arun