We have built a new Redhat Linux server running Oracle 10g for our Baan environment.

Using bdbpost6.1 to move data from our current system is proving to be difficult in terms of performance. We have one data load running now which has been doing so for over 48 hours. The company isn't particularly large but does have logical table links to another company. If I run the ttaad4222m000 - Count Number of Records by Table a figure of 900000 is reported.

Steps taken to try and improve performance.

bdbpost6.1 - Create rows before index set
db_resource - max array and fetch set to 100, first_rows_hint:2
Porting Set - Latest 6.1c.07.16

When you look at the server resources, be that from Linux or Oracle monitoring tools, the system is not being over worked at all. It's a new server with Quad core processor, 8mb RAM running 64bit Redhat. The problem almost appears to be that Oracle has reached a limit and is throttling itself back.

Can anybody suggest other areas for me to look at in an attempt to improve performance?

Thanks

PJ

I ran into that too. It was related to 10gR2. bdbpost was fast in 10gR1

I applied the latest 10gR2 patch and did some db tuning and that helped significantly, but i haven't fully revisisted it yet, i'm still on 10gR1 in production.

Dave

We've loaded the latest patch sets on 10gR2. What db tuning did you do?

PJ

In any export/import situation, the server will be not loaded CPU wise. How can a single bdbpost process load a server with several (4) CPU's?

If your disk subsystem is up to it, start several exp/imp or bdbpost/pre in parallel.

Regards

Markus

We've loaded the latest patch sets on 10gR2. What db tuning did you do?

PJ


first is you need to turn off the recyclebin
alter system set "_recyclebin"=false scope=both;
// if that doesn't work, do scope=spfile

That will probably fix it.

The rest was tuning the pga/sga size and tuning my LMTs

I am also a big advocate of the tuning suggestions in solution 166049
on the baan support site.

Dave

In any export/import situation, the server will be not loaded CPU wise. How can a single bdbpost process load a server with several (4) CPU's?

If your disk subsystem is up to it, start several exp/imp or bdbpost/pre in parallel.

Regards

Markus

That's true, but with 10g on linux it was really bad. 9i i could import in like an hour, 10g it went to like 24+ until it was tuned.

Dave