Test Deployment: Deplpoy Sandbox/DEV/QA/STAGE on low hardware spec for testing

KNIME Business Hub
1

Hi,

on the Spring Summit in Berlin I exchanged with a few attendees about HUB deployments. We realized that albeit CD is envisioned in the docs, the ability to deploy on servers with limited ressources to test before PROD-Deployment seems missing.

I vetted the BYOL option in AWS and the available server sizes are predefined.

Screenshot 2025-04-14 205644
Screenshot 2025-04-14 205644862×1216 69.8 KB

Henceforth, in order to test deployments without incurring high hosting (or license fees), I’d like to propose to allow Sandbox/DEV/QA/STAGE Deployments on low hardware spec environments.

That would balance the need of a Continuous Deployment Pipeline but also saveguard KNIMEs financial interests since the deployed HUB is barely usable if usable at all on a low spec server.

Best
Mike

2

Hi Mike,

Thanks for reaching out and for sharing your thoughts.
At the moment, it’s possible to use separate Execution Contexts (ECs) within a single Hub installation to support different stages like DEV, QA, etc. Alternatively, setting up an additional Hub environment is also an option if complete separation is required.
Regarding hardware specs, the minimum requirements for Hub are already quite low. Even with limited usage, the core services still need those resources to function properly. The main area where you might reduce resource usage is on the execution vCores.
We’re also curious to better understand your specific need for separate installations. Would you mind sharing more about the reasons behind that setup?

Thank you,
Attila

3

Hi @ACseresnyes,

the proposal is less about right sizing but being able to test updating a Hub deployment without inflicting too much costs. To test a deployment it is not necessary to have 16 vCores and 64 GB of RAM.

At present the smallest instance adds an additional $-US 560 (on demand / Month) plus the Hub License fee and an assessment cannot be carried out so fast.

Being able to deploy on a m5.xlarge (4 vCPus, 16 GB Memory) should still be sufficient while not allowing a production-like and therefore misuse of the license.

Best
Mike