Re: AQ streams expected performance in 10gr2

This isn't completely unexpected - you can expect the streams subscriber processes to be more labourintensive  than the "end-user" driving processes, and possibly doubling the actual number of transactions through the background activity.

But it is hard to say whether that level of transactions is good, fair, or bad for your hardware without some details of what a 'transaction' is.

If you haven't tried it so far, try using statspack at level 7 to find out where most of the work is going. It may give you some clues. Look, in particular, as the SQL to see if some 'unknown' SQL is actually responsible for most of the CPU time, or whether it is something about your application SQL.

Run your processes for at least few minutes, and take a pair of snapshots that eliminates the first and last minutes of the test.

-- 
Regards

Jonathan Lewis
http://www.oracle.com/technology/community/oracle_ace/ace1.html#lewis

The Co-operative Oracle Users' FAQ
http://www.jlcomp.demon.co.uk/faq/ind_faq.html

Cost Based Oracle: Fundamentals
http://www.jlcomp.demon.co.uk/cbo_book/ind_book.html


<gmendonc_at_gmail.com> wrote in message 
news:1146684978.641197.129180_at_j33g2000cwa.googlegroups.com...


> We are testing using AQ streams with JMS as the consumer on 10.2.0.2.0.

> We are having performance problems on the dequeue side. We are

> currently able to run approximately 2000 transactions per second on an

> hp/ux 3440 box with 4 cpu's and 8G of memory. After that the cpu's

> become a bottleneck. Both the enqueing and dequeing are on the same

> box. We have 1 enque table with 30 threads(processes) and a trigger to

> populate an intermediate queue table, we have 10 deque tables with 20

> threads (processes). We have tried a few of the performance tricks from

> the docs but nothing seems to be able to reduce the amount of cpu that

> gets used. Has anyone experienced something similar? Any suggestions

> would be welcome.  The sga is configured with 2.6G of memory, we've

> tried bumping it up to 4 but the box seems to use more swapfile then

> with little gain in AQ performance.

>

> Thanks,

> George

>