Info binary load operation

[gianluca89]

Hello

For our solution, we set up three following layers based on 3 Qlik apps:

App ETL1 - Reading data from the feeding source and generating QVDs.
App ETL2 - Reading QVDs from app ETL1 for data model generation
App ETL3 - Visual presentation layer for the end user.

The end user expressed the need to enrich the ETL3 app with additional data sources (via excel file import) in order to gain additional business information.

To enable this functionality, we are considering to expose a duplicate of the app generated in ETL3. We will call this new duplicated app 'ETL4'.
We want to retrieve in ETL4 only some data of interest belonging to the ETL3 app via the 'binary load' function.
At this point our doubt arises:
does the binary load operation duplicate the QVD space in ram due to a necessary reload? Or can we use some mechanism to reuse the QVD already generated in the ETL3 app?

Thank you in advance for you support.

Best regards,

Gianluca

[Or · Accepted solution]

QVDs do not take up RAM. They are files and thus take up space on your file system (hard drive / network drive / whatever), and using Binary Load does not impact QVDs in any way.

However, if both the ETL3 and ETL4 apps are in use, each will be individually loaded into the Qlik Server memory (or local memory if using Desktop) and each will take up RAM individually. There is no way around this that I'm aware of since they're individual apps even if they happen to be loaded from the same underlying source.

[marcus_sommer]

Personally I wouldn't tend for an ETL4 layer but it will depend on the entire data-architecture and the particular requirements if this kind of overhead may still be the most suitable approach.

The additionally data may just be loaded/transformed within the ETL1 layer and then loaded through the whole chain. Another approach might be to add this load to the binary - the binary must just be the first statement but afterwards you could load more data and/or remove/transforming them further.

Also thinkable would be to include these data per direct discovery.

- Marcus