Filip,
1) Performing sub-setting as part of the very first VDB creation is not supported for HANA at this time. As HANA and our product evolve together, it might come. For now, you should create a "golden" VDB first which is a full dataset. Then, transform it anyway you want, including subsetting, masking, adding synthetic data, etc. Now the magic...create VDB copies of the golden VDB and distribute to teams. We call this "transform once, distribute many."
2) I'm not a HANA expert, but my understanding is that you can use less RAM than the full size, but it's risky. It's not a best practice, and your HANA system will be unstable, but this might work OK for you in non-prod systems.
Again, I'm not a HANA expert, but for most databases once you have a working system you could do non-subsetting techniques to make it smaller, such as dropping tables, schemas, tablespaces, or data files. Obviously, your app needs to be OK with that and HANA must support that. Anything you do which makes the data set smaller will reduce your RAM needs.
3) Masking of HANA is not supported out of the box at this time:
https://maskingdocs.delphix.com/Getting_Started/Data_Source_Support/#support-matrix This matrix does evolve, and it may be supported at a future date.
For now, you must provide your own driver, and the support model is described here for Extensible Masking Connectors:
https://support.delphix.com/Delphix_Masking_Engine/Delphix_Extensibility_Support_(KBA5571)I can share that we ourselves have masked HANA successfully with 3rd party drivers in our labs. It was quite similar to other databases.
Delphix Services can assist you with the initial install and deployment of HANA Masking with a 3rd party driver.
------------------------------
Ranzo Taylor
Technical Services Director
Delphix
------------------------------
Original Message:
Sent: 09-09-2021 11:57:33 PM
From: Filip Van Grembergen
Subject: SAP HANA memory requirement on target system
Hi Ranzo,
After discussion with the team, following clarification questions were raised:
1. Is it possible to refresh a sub-set of the production data from the source production system to the target test system? For example, it is possible to slice the data by company code, by time period (fiscal year), or other criteria to select the data that should be refreshed from source to target?
2. Are there are any other capabilities to reduce the target HANA system memory requirements compared to the source HANA system? Given that the production HANA system memory size (=RAM) is typically large (and hence the system is expensive), it would be useful to have the ability to reduce the target HANA test system RAM memory requirements (to reduce the cost of the test environment).
3. Is masking supported for SAP HANA? If so, are the masking capabilities the same as for Oracle?
Thanks,
Filip
------------------------------
Filip Van Grembergen
Community Member
SAP America, Inc.
Original Message:
Sent: 09-09-2021 02:08:36 PM
From: Ranzo Taylor
Subject: SAP HANA memory requirement on target system
Filip, your HANA memory requirements will not change.
While Delphix does reduce overall physical storage of multiple systems through compression and block sharing, each database will still see 1 TB of files.
For non HANA systems, we can sometimes turn down buffers because the Delphix Engine has it's own cache which can help to speed up some workloads. For HANA, I understand that it expects RAM to equal the total DB size--in your case 1 TB.
Ranzo
------------------------------
Ranzo Taylor
Technical Services Director
Delphix
Original Message:
Sent: 09-09-2021 01:32:09 PM
From: Filip Van Grembergen
Subject: SAP HANA memory requirement on target system
It is clear that Delphix results in storage footprint reduction on the target HANA system. What is not clear is whether this results in memory reduction as well. As an example: suppose a source HANA database of 1 TB RAM (memory). When creating the copy, will the target HANA system also require 1 TB RAM? This is important as the cost of an SAP HANA instance is primarily determined by the memory size; storage cost is significantly less important.
------------------------------
Filip Van Grembergen
Delphix Community Members
------------------------------