COSIMA Q2 2026 compute

COSIMA
1

Hi COSIMA,

Who wants additional ACCESS-NRI compute this quarter? Could you please outline your requested budget and a sentence on what it would be used for below. So far our approach has been to prioritise model development / community runs that will benefit multiple people on COSIMA compute and for people to try to use up university projects first for individual science projects. But open to discussion on this approach.
Tagging @dpath2o @pearseb @aekiss who all have requests.

Also, while we’re at it, it would be great to have some clarification about whether we should be using central ACCESS-NRI (tm70) or COSIMA working group (ol01) compute for development projects. Specifically this quarter we have requests for:

  • OM3-25km testing runs
  • final optimisation of WOMBAT-lite
  • sensitivity experiments with WOMBAT-mid

Would love your opinions on whether these should go on tm70 or ol01 @clairecarouge @cbull. People requesting seem to think they should be going on ol01, but my impression is that other working groups are mostly using their compute for science not development? Thanks!

2

@adele-morrison, thanks for the message. There are a few things to clear up.

  1. What can be run on the ACCESS-NRI’s Merit Allocation Scheme (MAS) allocation, i.e. on the working group projects?

Going back to the MAS guidelines, the first assessment criteria is described as:

Priority will be given to requests which contribute to the ongoing development of new ACCESS configurations, or to enhance existing configurations.

It does not speak of science vs technical tests. Any request for the development of new ACCESS configurations is considered.

The MAS guidelines do not dictate the type of ACCESS-related activities the community needs. The guidelines are to ensure the projects are aligned with ACCESS developments and of interest to a reasonable portion of the community.

  1. What are the Community Working Groups (CWG) typically using their SU for?

The majority of the working groups’ allocations has been going to science projects. However, almost the entire development of ACCESS-AM3 was done on the land working group project because we needed a place where NRI staff and CSIRO folks could work together. The same happened with the merging of historical branches for CABLE. Various projects have different requirements, so there isn’t a one-size-fits-all process, and we are not saying you should follow the AM3 process. This is an example to show that working group project allocations can and have been used for development purposes, not just science. The usage is guided by the community’s needs.

  1. What is tm70?

The NCI’s scheme for ACCESS-NRI resources covers both the MAS allocation and the ACCESS-NRI internal allocation. This was decided to simplify the transfers of resources and better optimise across the whole resource purchase. However, we only report on usage for the MAS allocation (the SU allocated to the working groups and the storage allocated to working groups and reference datasets) to NCRIS and the ACCESS-NRI’s board.

Within that framework, tm70 was designed as the ACCESS-NRI’s staff main project. Some teams have other projects for various purposes but tm70 is our staff default project.

  1. How to choose between tm70 and a working group project?

As a general rule, a non-ACCESS-NRI staff member should not have to use tm70 and should use a working group project instead. It is impractical for ACCESS-NRI to manage the requests from everyone and we are not best placed to assess the merits of use. Community needs are then best managed by the working group co-chairs. The co-chairs should then consult @jasmeen_kaur, myself and the other WGs co-chairs when their consolidated needs are above their allocation and they need extra resources.

There is a bit of flexibility on this, especially towards the end of the quarter, where changing a project string in a file can be faster than moving SUs. And there are surely some cases that really won’t fit within the rule of “tm70 for ACCESS-NRI staff only”, but these should be rare.

In conclusion, it seems @pearseb 's and @aekiss 's requests would fit under the working group project (ol01). We can up the allocation if required.

Let us know if you still have some questions.

@clairecarouge and @cbull

3

Thanks Claire.

Copying in the requests that were emailed to me:

@aekiss:

  • The 25km IAF run. Will cost ~1MSU. (although has most of this already been run on ANU compute now Andrew?)

@pearseb:

  • 200 KSU for final optimization of WOMBAT-lite
  • 1 MSU for sensitivity experiments with WOMBAT-mid

Plus we (@adele-morrison, @claireyung, @Wilton_Aguiar, @PSpence etc) would like to do a new spinup (30 years) of the ACCESS-OM3-panan-iceshelf config with some of the recent GFDL and ACCESS-OM3 bug fixes and improved melt tuning:

  • 7.5 MSU (though this may take a while to setup, and we likely won’t get to this until the last month of the quarter, so could likely only run ~3 MSU in the time left this quarter).

How about @aekiss and @pearseb you go for it now. And @clairecarouge perhaps if there is extra available, then we could have some of that to run the ice shelf spinup from the start of June?

4

Thanks Adele I’ll get stuck in. I’ll monitor and report back what I’ve used.
pearse

5

Thanks @adele-morrison . I’ll bump up the COSIMA allocation to 5 MSU now and we’ll see later in the quarter when more is needed. Considering the time of the day, the new allocation might only appear on Wednesday.

6

I’m running another 25km IAF OM3 test on ol01 - will take up to 700KSU - hope that’s ok?