Can I ask a RAC user a favour?

Anyone with access to a multi-node RAC could run this test for me?

1- Create a sequence seqname with default (20) cache. 2- Write one small PL/SQL block that loops 10000 times on: select seqname.nextval into l_seq from dual; insert into some_table values(l_seq,something_else...); 3- Write one small PL/SQL block that loops 10000 times on: insert into some_table values (seqname.nextval,something_else...); 4- Run 2 and 3 in two separate SQL_TRACE=TRUE sessions, one after another.
5- Repeat 4 with the sequence altered to cache=200. 6- check the tkprofs for major differences in stats and I/O?

I'm getting something unexpected (to me): in 2, the overall and individual SQL stats look the same with or without the larger seq cache.
in 3, they are very different as the manual would lead me to expect.

I don't understand why the select from dual is not showing the extra I/O and stats I see on the "insert ... values (...nextval...)", when the seq cache is small.

Trying to figure out if it is just my deodorant or if this is something I should have read somewhere and be aware of.

Cheers and TIA
Nuno Souto
nsouto_at_bizmail.com.au

--
http://www.freelists.org/webpage/oracle-l