Solaris/SAP Memory Analysis/Audit

[Former Member]

Experts,

We have a landscape that is only a couple of years old, but seeing some paging and buffer issues. We are running in a solaris/Oracle environment, and I was wanting to do some memory analysis, or a memory audit to make sure that we had everything configured correctly.

Memory is not really my stong suit, so if any of you have some handy resources that you dont mind sharing, I would like to research and better understand.

Our Prod Landscape is this:

Sun e2900 (#1) 8cpu (multicore) 64g

-

---

Global Zone - Ecc 600 ASCS /Oracle 10.2 Cluster (Slave)

Local zone - BI 7 CI+DB

Local Zone - XI 7 CI+DB

Sun e2900 (#2) 8cpu (multicore) 64g

-

---

Global Zone - Ecc 600 ASCS /Oracle 10.2 Cluster (Master)

Local zone - ECC CI

Qty 3 - Sun V490 (#1, #2, #3) 4cpu (Multicore) 16g

-

---

Global Zone - OS only

Local zone - ECC APP

Local zone - BI APP

Local zone - XI APP

I have the performance tuning book, and have been studying up on buffers and parameters. But first mainly looking to research how memory is assigned, where, and determine if any changes need to be made, or there is memory room to increase buffers.

Thank you.

Phillip

[Former Member · Accepted solution]

Your question is not clear. But go through ST02 for various buffers and swaps. If swaps are more than 10K then you may need to look at tuning those buffers. Make sure you have enough memory in the systems before tuning these parameters.

Have a look at

[http://help.sap.com/saphelp_nw04/helpdata/en/02/96253c538111d1891b0000e8322f96/frameset.htm]

[Former Member · Accepted solution]

Apologies. I was struggling with how exactly to articulate the question. And even more with how to go about determining where we are with our memory both inside and outside of SAP. With Solaris you have Zones (containers), projects, oracle, sap, user environment settings, etc. All of which have an impact on memory, either from consumption, or limitation of memory (projects and user environment settings.)

I have looked at ST02, and we do see swaps of greater that 10k, however, before increasing the buffers, I'm trying to determine what memory we have available, and where the current memory is being used.

Our response times are not absurd. We hover around 800 - 1000 ms. Our CPU is average less than 15% util. But I think there is room for improvement/tuning on our memory and buffers, because performance does degrade over a period of weeks without an app restart.

Thank you for your response, and the help link. I am reviewing it now.

Best Regards,

Phil May

Edited by: Phil May on Sep 11, 2009 9:52 AM

[Former Member · Accepted solution]

That was a very helpful link. Thank you. Points awarded.

[Former Member · Accepted solution]

Hi,

Sometimes it also happens like we have already assigned a high value to the PXA buffer but still swaps are reproducing. Here we could suspect of having a PXA fragmentation and more on this you can see in sap note 1267828 Swaps in program buffer even though plenty of space is free.

Do you get more dump in ST22 related to PXA like PXA_NO_SHARED_MEMORY and please check the dev trace file of PXA INITIALIZATION wther correctly getting initialised or not.

Thank you,

Tilak