13 line item dimension cube into BWA?

[Former Member]

I have heard earlier that BWA does not support line item dimensions due to swap of data between blades. not sure this is still an issue.

My cube has 13 line item dimensions. For performane reasons, we created 13 line item dimensions as cube storing 240 million records and user want to analuse all this data.

I want to push this line item cube to BWA for better performance and like to know the impact.

thanks for advance replies.

Thanks

[former_member93896 · Accepted solution]

Hello "Texas User",

let's set a few things straight:

1) There are no limitations using BWA for InfoCubes with line-item dimensions.

2) In general, WebI 3.x as well as Webi 4.x work well with BW and if you index your data with BWA, the queries will run faster.

3) SAP is continuously improving and optimizing the integration of BusinessObjects and SAP NetWeaver BW. Therefore, BI 4.0, BW 7.3, and BWA 7.20 "run better" than older releases. This is what every customer expects so there should be no surprise if SAP recommends to look at the most current release.

Looking at your specific case, yes, you can certainly index your InfoCube and make all data available to your users for ad-hoc analysis. For Webi, please make sure that the "query stripping" setting is turned on (otherwise way to much data will be read).

There is one critical issue that you must be aware of and educate your users about: The more dimensions a user add into his query drill-down (without proper filtering), the more data will be selected and the less effective BW Accelerator will be in speeding up the query. In your scenario, the worst case would be that a users creates a report with all 13 dimensions in the drill-down resulting in the selection of all 240 million records. These kind of queries will certainly exceed system resource limitations (memory and CPU) and will abort. It is essential that you tell your users not to behave in such a way. Filter your data sets to avoid such large data selections.

Regards,

Marc

SAP Techology RIG

[Former Member]

Hi,

Agreed that BWA is extreme powerful but just try using the aggregation property "Before Aggregation " in any of the queries and the query will stop hitting the BWA. The reason being the data is read from the cube at a record level and indexes are formed at an aggregated level.

But in your case the query will hit BWA and also there will be performance improvement, no doubt in that, but how much of it will be visible is again dependent on the data volume.

For sure you must go with the BWA for this requirement it will help.

Regards,

Shyam.

[Former Member]

Hi,

The BWA has a powerful search engine and it will improve the performance for sure, but the % of improvement will depend on the number of records you are accessing. If you are reading at a single record level then BWA wont help much.

Regards,

Shyam.

[Former Member]

My assumption for BWA has no limitations. That's why we spent so much for BWA right?

It should support all navigation included single record reading to whole set of data.

I have 300 million of records in cube. We don't know what kind of adhoc analysis user planning to do.

We put all 14 fields from base cubes in query allowing user to do anything with data to make better decisions.

My question is making 13 line item dimension in cube and putting 300 millions records in BWA improve performance or not?

Space or cost is not concern here. Do we able to do it?

Does Web intelligence query come back? Or die?

[Vitaliy-R]

My assumption for BWA has no limitations. That's why we spent so much for BWA right?

Rewording Mark Victor Hansen: "Only in imagination, there's no limitation".

Best cases for BWA performance are documented; unfortunately the only link I have on hand is to BIExpert article [Tips and Techniques for Optimal Query Performance with SAP NetWeaver BI Accelerator|http://www.bw-biexpertonline.com/article.cfm?id=3665].

> It should support all navigation included single record reading to whole set of data. 
> I have 300 million of records in cube. We don't know what kind of adhoc analysis user planning to do. 
> We put all 14 fields from base cubes in query allowing user to do anything with data to make better decisions. 
> My question is making 13 line item dimension in cube and putting 300 millions records in BWA improve performance or not? 
> Space or cost is not concern here. Do we able to do it? 
> Does Web intelligence query come back? Or die?

So far there is nothing worrisome in your description. 300M recs is not mind-blowing. The areas of watch-out:

1/ There used to be performance implications with line-item dimensions in the past, when number of records in dimension was higher than number of records in fact table. But it was 2-3 years ago...

2/ Once you put something on top of BW, like BO in your case - you need to check where time is spent, because in the past communication between BI tool and BW used to have impact as well. Make sure you have latest patches there.

Cheers,

-Vitaliy

[Former Member]

Hi Vitaliy & Texas

Yes we had issues in the past with line item dimensions. If I remember correctly they were using only a single blade-not distributed, therefore no parallelization-, but with BWA 7.2 i think the indexes in the above case will be created as FLAT index, therefore assuming that it will work the way it supposed to be. But I am not certain, can you comment on this or Marc - I haven't heard from him yet

When it comes to the Webi performance SAP is emphasizing that it will go away with BOBJ 4.0 which I agree ( it will use BICS instead of MDX). So the BWA solution itself may not be the answer for you

Cheers

[Former Member]

Hello All,

Then what is point of selling product BOBJ 3.x which is unusbale and we had to wait for next version 4.0?

It has been advertised by SAP that Web Inteligence answers adhoc analysis questions so far and users are in such a frustration to understand WebI does not support detail level reporting.

I have no clue that 4.0 will fix this issue..?

Thanks

[Vitaliy-R]

> Then what is point of selling product BOBJ 3.x which is unusbale and we had to wait for next version 4.0?

To make revenue? Not that I appreciate that, but I think that's the point in every company, especially public one.

[Former Member]

Hello,

BI 4.0 will leverage the native BW connectivity to eliminate certain levels, hence increase the speed for Webi.

With this change, your Webi report and BW query will have a very similar run time - currently Webi is considerably slower -.

Now we can talk about BWA, as you know today it will only increase your database read time; not the olap, network traffic, communication etc. If the DB time is a small percentage in this overall picture (from BW query to Universe to Webi ) you won't gain a lot, that's why I brought up the BO 3.x Webi performance issues to this discussion

Cheers

Tansu

[Former Member]

Hello,

If you have space on BWA why don't you index this cube and run couple reports to see if it works or not. If the performance doesn't improve just delete the cube. Or find out some key queries from this cube, run them and check the statistics. If DBSEL and DBTRANS are almost same I would think that you won't see any performance improvement.

To answer your question, line items do work with BWA. However it's really depends on how the queries are built. If you select large number of records from the line item dimension(no aggregation) then you won't see any improvement.

Cheers

Tansu

[Former Member]

Hello Tansu,

Thanks for reply.

This cube is used in Web Intelligence, we included all info objects in the query for adhoc analysis.

There is no specific bex layout, user can use any 14 fields.

We want to make 13 line items and use in webi.

Can we do that?