Program buffer

[Former Member]

Dear all,

We are using R/3 4.6C on HP-UX 11.0 with Oracle 9.2.0.5 and kernel sup. package 2039. On this server we have two production R/3 instances. Server has 8 CPUs, and 8GB RAM, with defined swap of 20GB.

Can You PLS explain:

- we have more then 92000 swaps on the program buffer in one week only on one instance which has more users and more transport activities;

- Should I increase the parametar abap/buffersize (currently 500000kB)? And how much should it be?

- ipc/shm_psize_06 should it be = -10 and ipc/shm_psize_02 = -10?

I'm not sure what are the steps to increase the program buffer, and how much should it be increased?

Thank you in advance.

Regards,

Ruzica

[Former Member · Accepted solution]

Hi Ruzica,

Have you seen OSS note 103747 which describes the recommended parameter settings for 40B and above ?

It looks like you may need to increase it slightly (profile parameter abap/buffersize). The note suggests that the system can tolerate 1,000 swaps per day but you are seeing over 10,000 per day.

In our busiest SAP system, R/3 Retail, we have a PXA buffer size of 600Mb and we experience between 1000-2000 swaps per day. However, it depends what mix of modules you are using in the system. This greatly influences the programs that are loaded.

You mentioned that you have two instances on the one server. Is this the CI and one application server for the SAP system or is it two separate SAP systems (i.e. separate SIDs) ?

If the two servers are part of the same SAP system, it may be worth looking at how the users are being distributed across the two servers.

Regards,

Gary

[Former Member · Accepted solution]

Hi Gary,

We have two separate R/3 instances and DBs on one server.

I have looked on the note 103747, but I'm still not sure how much and what should I do to increase the program buffer.

If I change the abap/buffersize, currently 500MB, to exp: 600MB, what should be else changed or checked, ipc/shm_psize_06 = -10 and ipc/shm_psize_02 = -10?

Thanks,

Regards,

Ruzica

[Former Member]

Hi Ruzica,

The specification of profile parameter ipc/shm_psize_06 = -10 is relocating the PXA buffer into shared memory pool 10. If you increase the size of the PXA buffer via profile parameter abap/buffersize, you need to ensure that there is sufficient space in shared memory pool 10 to accomodate this increase.

You can check the current pool usage via ST02->Detail analysis menu->Storage->Shared memory detail. Here you can see the amount of free memory in pool 10. You should also see the ABAP program buffer has been relocated into pool 10.

Once you have changed the profile parameter in RZ10, you can check the new memory allocation using O/S level program sappfpar, as follows:

sappfpar check pf=<instance_profile>

Regards,

Gary