Performance Degrade in Volume Test- BW, BPC, BPS, BW-IP

[Former Member]

In the volume test of BW/BPC/BPS/BW-IP(BW 7.40) solution, you may see performance degrade in volume test compared to single user execution. Or performance is good in the first beginning but degrade when the system has been used for a while. Here is the steps how to narrow down the root cause when issue is related to OLAP cache.

 

First of all, make sure there are enough resources in your system, like number of work processes, number of RFC WP configurations,
number of CPU and Memory for your volume test. If hardware utilization reaches to 100%, there could be a lot other unexpected errors or performance degrades caused by hanging processes.

 

 

If your system has enough resources and in transaction SM66 or SM50 you find most work processes hanging with report CL_RSR_CACHE_STORE. The reason could be the OLAP cache in the system.

 



In ST12 trace, you can find most of the runtime is spent on function CL_RSR_CACHE_STORE============CP.



After you delete the cache from transaction RSRCACHE, you may find the performance is much better.



Check BW statistics table RSDDSTAT_OLAP for logs, you can find most of the expensive event id is 2505.

 



 

Event 2505: Read Cache Entries

This event includes the search of a storage entry or data entry and the reading from the persistent medium if this has not yet happened (SPs that have already been read are buffered locally in the roll area)

 

Here are some FAQ SAP Note for OLAP Cache:

1805930 - New OLAP Cache function and parameterization

OLAP: Cache Monitor - Business Intelligence - SAP Library

OLAP Cache - SAP NetWeaver Business Warehouse - SCN Wiki

 

 

You can do the testing without OLAP cache to narrow down the root cause:

Disable OLAP CACHE globally in transaction RSCACHE.



 

OR

 

Turn off OLAP cache for specific query. For BPC or other planning query, the default query is  <Multiprovider name>/!!1<Multiprovider name>. Change the Cache mode from D "Cache in database" to "Cache is inactive"



 

You can even turn off internal OLAP cache buffer.

 

2017248 - Disabling internal OLAP cache buffers (local cache)

 

If your new volume testing shows that performance is consistent compared to single user execution or first time execution, then you needs to check out SAP Note 2117986 - OLAP cache read accesses very slow.

 

Or as a temp workaround, you can keep OLAP cache disabled if this change does not impact system performance too much.