Hi Syed,

 > We create a new partition every month. Whenever there is

> a new partition created and the job takes long time than the usual

> time, only the first day,

some questions:

* do you collect statistics on the partition level?
* do some of your statements use the partition level statistics (i.e. access exactly one partition)
* is the partition pruning based on a bind variable?

If the above points match you may try not to collect the statistics for the new (empty) partition, but to fake them (e.g. copy them from the previous month).

I observed similar case.  Drop me  a note if the theory is right, I can provide further detail.

Regards,

Jaromir