When do we have Push down code in SAP Hana

Question

I need an answer or better an explanation between SQL (Open/Native), CDS and AMDP.
I understand that in order to follow the rules of SAP and push the code down in DB HANA we must use CDS or/and AMDP. I have thought that if we still use only SQL Queries is the old way with which we use the Code in AS. I have read some articles or I have seen some videos in the internet and they have confused me.
Can someone explain me which is the best way for following the PUSH DOWN the CODE in HANA?

 - Use of SQL queries Open or Native what it is? 
 - Use of CDS or AMDP are PUSH DOWN CODE technique for Hana.

Thanks

Answer 1

“Code push down” means that you execute the expensive main part of your computations in the database, not in the application code (= ABAP). How exactly you do that is related only indirectly.

For example, instead of SELECTing from two database tables and then mixing up rows in the ABAP code, pushing down means you run a JOIN in the database. This reduces the number of round trips with the database, exploits the columnar characteristics of SAP HANA, and benefits from other code-near-the-data effects, such as reduced amount of data that needs to be transported to the upper layer.

Whether you implement that as an OpenSQL query, a classical database view, a CDS view, or inside an AMDP, is not the main question. We found that resorting to simple OpenSQL queries on CDS views is the optimal choice for standard cases. When things get more complicated, my answer to this other question may give some more guidance when to use what.

Disclaimer: although I am working for SAP, there are many other opinions on the topic and some may look at this differently, so kindly do not understand this as a reliable official guidance that fits everything.

Note that code push down is not a silver bullet that fixes everything. In scenarios with many parallel consumers, pushing down large operations may clog up the database for the other users, such that you need to resort to different patterns, or find ways to limit resource consumption.