Page tree
Skip to end of metadata
Go to start of metadata

As discussed in Convex Hull, even in case the HANA Analytics-API is used, it may happen(in rather rare and uncommon cases) that the convex hull reduces the data volume transferred from the database to the Analytic Engine. See note 2306646 and the following simple example(which should provide an idea how such a special situation can involve) .

Sample Query

Our very simple query uses 12 restricted key figures(regarding 0CALMONTH) where a kind of cumulated display takes place(see also example of note 1858333).

  • 1st KF: restricted to May
  • 2dn KF: restricted to interval May-June
  • 3rd KF: restricted to interval May-July
  • .... and so on

The characteristic Customer is in the drilldown and there are only two data records in the provider which fulfill the query restrictions:

  • Customer Number 220: key figure value 10588 DM for month 05.1999
  • Customer Number 40: key figure value 10588 DM for month 05.1999

Since these data records are relevant for all restricted key figures we get the rather exotic situation where the usage of the HANA Analytics-API increases the data volume:

Query Definition

Checking Query in RSRT

We run the query in Transaction RSRT and use the button Execute&Debug to get the trex statement displayed:

Internally the following data records are transferred from the database to the Analytic Engine:

Although there is only one relevant data record for the customer 40, three records are transferred since the information about the relevant FEMS((the FEMS is basically the BW-internal representation of the filter conditions for a restricted key figure) is delivered as well. Actually there would be 12 records(12 restriced KFs, 12 FEMSes), but the feature fems compression reduces the number to 4. E.g. the first record is relevant(assigned to) for the FEMS 5,6,7 and 8, the seconds one for FEMS 1,2,3 and 4.

Query Result

SQL used

In case we switch off the usage of the trex API(possible in this case), SQL is used to retrieve the data from the database. Only 2 records(1 per customer) are transferred to the Analytic Engine. So, in comparison to above, just 1/3 of the data volume has to be processed in OLAP. You can imagine that in 'real' scenarios this might be a decisive difference.

We run again the query in Transaction RSRT and use the button Execute&Debug to switch to SQL(bei choosing trexops=0, see Operations in BWA/HANA) and get the statement displayed:

SQL statement:

When we check the records transferred from the database we can see that there are only 2 records:

Convex Hull used

If we use the RSADMIN parameter introduced in note 2306646 to force the usage of convex hull(only possible with a trick in this simple case), the trex statement changes to

There are no local filters(FEMS>0) any more and the number of records reduces to two:

  • No labels