New interface to SAP BW

[Renee]

Hello,

we just testing TM1 951for planning purposes and we try to load hierachys and data from SAP BW 7.0 via sap querys. We are using the "new" IBM interface and not syskoplan,

We have a lot of problems e.g.

elements of a hierachy gone lost
it takes more than 45 minutes to load a hierachy with 6 level and 15000 elements
no data reaching TM1

Does anyone have some experiences with this interface. Do you have suggestions for the design of the sap query or anything else?

Thanks in advance
Renee

[Klopse]

Hello Renee,
yes we are facing the same issues. Poor performance for dimension load. There are no benefits from the multithread feature. It depends on the used technology. Relax and get some coffee

If You have the chance to use a alternative SAP Interface and You can access a relational staging database You should have a look on Theobald or Cubeware, easy of use and very fast interfaces for BW and R/3. Theobald (http://www.theobald-software.com/en/pro ... ractis.htm) works without RFC implementation and is available for 4000 €. Cubeware (http://en.cubeware.de/products/cubeware ... orsap.html) supports direct load to TM1.

Here are some tips from the TM1 Support if You want to continue with this . Please note, You have to design special queries if You want to use the Multi query / thread load feature. The Package Connector does not work without this feature Alle dimensions in the row, no one in header, one small and dense dimension in the column with max 20-30 elements. This dimension will be declared as variable for iteration by the interface.

Regards

Klopse
<
"Tips and Tricks with the TM1 Package Connector

Large Hierarchies and data extractions

Ensure you have a cogtr.xml file for the TM1 Package Connector containing
the following:

<?xml version="1.0" encoding="UTF-8"?>
<Settings>
<Sections>
<Section Name="Transformer">
<Preference Name="SuppressNull" Value="1"/>
<Preference Name="SegmenterParallelQueryCount" Value="4"/>
</Section>
</Sections>
</Settings>


The SuppressNull preference value of 1 means do not suppress NULL rows.
This means that hierarchy queries will not include the NON EMPTY clause and
be much faster. Otherwise, only members with fact data are returned,
causing a read against the fact data of the cube, which can be quite slow.
Alternatively, you can select the "Edit Governors" menu item in Framework
Manager and turn off NULL suppression for SAP BW.


The SegmenterParallelQueryCount determines the number of parallel queries
executed by TM1 Package Connector. I have seen performance benefits even up
to 32 parallel queries, but there are diminishing returns as you increase
the number. Try a lower number and check the impact on their SAP BW server.
For reference, I did an extraction test on a cube with about 2 million rows
of fact data. Using a value of 4 was about 3x faster than serial queries,
and using a value of 16 was about 5x faster.

Ensure you build a BEx query that includes a restriction suitable for
segmenting the requests. I have found that a time-based one works well,
like by month.Then the query results per chunk are of a similar size.
Ensure to do "Edit Prompts" and set the segmenter to this prompt value so
that the stream BAPI processing occurs. And ensure you set the cube build
process to "Accumulate values" since with stream BAPI you get rows out of
the fact table and values may not be aggregated." >