Combining LARGE Datasets in Calculation Views

[Former Member]

Dear HANA Modelers

I have recently started building calculation views and one thing I learned is that when working with large datasets we have some performance issues

For example there was a need to link, Orders , Deliveries and Billing documents together and each of these are rather large datasets on their own

I first created analytical views for each of ORDER, DELIVERY AND BILLING and then built a CALCULATION VIEW that joins them all together

What we learned is that running it for all the records i.e getting aggregated information for some entity across the entire data set without filtering was just not an option due to performance

I had to restrict the END USER to enter a customer number and or sales doc no in their WEBI report that is built on top of this view so that it can run one customer at a time providing them a history of the above 3 entities for a given customer

IN a similar exercise I tried to tie TWO very large datasets of about 800 million rows and that also performed very poorly

Have folks worked with LARGE data sets joined via calculation view and if yes can you please share your thoughts on the topic

Regards

Message was edited by: Tom Flanagan

[nikhil_joy2]

Hi Aravind,

As mentioned here already, are you pushing the filtering operation to the lowest level. PFB a nice document from which shows the effect of pushing down filtering operation to lowest possible level.

http://scn.sap.com/docs/DOC-41562

Hope it helps.

Regards,

Nikhil Joy

[Former Member]

Hi Arvind,

There are a few standard guidelines like

• Selecting the only necessary columns for the report.
• Calculation after aggregation
• Filtering the data at lowest possible level in the information model

Other optimizations in the model might also be performed as each case would be different.

If these options are also not sufficing the need then this report could be an apt use case for BW on HANA, rather than just HANA.

--

Shreepad

[rama_shankar3]

That's a good write-up from Ravi that Nikhil referred to. Good luck!