|
How much effort is required to deploy your own REF environment? |
Answered by
Darwyn72
May 28, 2026
Replies: 1 comment
|
This depends on how you want to run it. For example, on ESGF they use Kubernetes as it is well supported although it would need a Dev-Ops team to implement in your institute. The REF does not require huge amounts of compute; hundreds not thousands of core hours. Some modelling centres are running local copies on their own HPC’s but this can require work-arounds as each HPC system is different. Currently, the API’s are pointing to the AFT REF version, so there would need to be some discussion of requirements. |
0 replies
Answer selected by
Darwyn72
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
This depends on how you want to run it. For example, on ESGF they use Kubernetes as it is well supported although it would need a Dev-Ops team to implement in your institute. The REF does not require huge amounts of compute; hundreds not thousands of core hours. Some modelling centres are running local copies on their own HPC’s but this can require work-arounds as each HPC system is different. Currently, the API’s are pointing to the AFT REF version, so there would need to be some discussion of requirements.