Minutes based on the minutes that Catarina sent by mail.

There minor additions and there might be errors introduced by  me (Markus) when reformatting the minutes.. 

 Discussion on EGI/GreenDigit - WLCG sustainability forum

WLCG Sustainability forum intro and goals

The following is a TLDR from Markus's minutes on the agenda https://indico.cern.ch/event/1557640/ interspersed with comments from the meeting

• Starting point: actions from WLCG/HSF workshop

WLCG Sustainability - Next Steps

▾ Sites:

▾ Benchmarking of compute, storage and networking 

▾ Optimisation of storage and compute: Event generator output shared across experiments: meeting in the EvGen / Sim Forum [link to be sent later]

• Tommaso: possibly difficult,

1.  the optimisation of generators is slightly different (e.g. PDFs/total cross section)
2. is combination going to be more difficult if we share events so they aren't statistically independent?

CD: when is the option not to do what we're doing right now? Try to think about HL-LHC and statistical uncertainties.

MS: Experiments are freezing things 2 years before the start of HL-LHC, therefore there is little time.

CD: Good time to start seriously in the next year. 

DB: Decision made by experiments in any case.

What is WLCG contributing to?

Metric to encourage them to make this decision. If you do this, then you motivate to optimise.

Move from site-based power accounting to job-based power accounting through benchmarking suites.

Metrics:

• Physics per consumed units of energy (number of events generated/analysed) <--- this means covering the full process, counting an event only once, even if reconstruction/calibration is run n times.
  • Difficult to agree on, could be done for EvGen/G4 in a straightforward way 
  • This could be an example for future (harder) things
• Encourage people to use fast simulation / FlashSim that can be shared with other experiments  <-- needs a study project, retrain the base model for another experiment and retrain
• Relative improvement of an experiment wrt itself, rather than each other
• Highlight areas where they can share more (see above)
• Also WLCG can give information about what gains can be made by the experiments. 
• Adoption of heterogeneous architectures   <-- maybe for more than is done now ( HLTs, event generators (very soon), a bit of ML training (currently small)
• How to support and benchmark new technologies  <--- HEPScore for GPUs is being worked on, but progress is slow.... 

How do we get the forum organised?

▾Where do we discuss, how do we keep track and identify additional topics

Topical meetings, short (1h) and focused on one subject only to have an in-depth discussion.

Working / technical meetings 

Who is discussing (identifying core group of people)  <----- action required, we need to identify candidates. 

"Coordinators"/organisers: MS (CERN, T0), CD (software/liaison with HSF), DB (WLCG and non-CERN sites)

Mandate: ensure that progress happens, invite for talks, setting clear goals and areas where work is needed (building an agenda). Also liaise with TCB.

Lobby: twice a year meeting of compute resource scrutiny group, looking at how "reasonable" use of resources can be asked for and needs for next round. Does not look into efficiency, failure rates…not in terms of CO2 impact. Try to get them to add CO2 consumption to the report, to raise awareness and communication to funding agencies that are sensitive to this.

Injecting environmental sustainability into HSF seminars (e.g. by inviting someone)

Core group of people who turn up at the meetings, loosely put together forum but have someone coming in from the experiments

Need an egroup + advertisement

Experiment liaisons

• LHCb Henrik G.
• ATLAS Zach
• CMS Liz / Phat?
• ALICE [DB will ask]

People who are interested in this topic

• Domenico, bringing link to HEPIX benchmarking
• Thomas Hartmann from DESY
• Glasgow / Manchester teams (Emanuele Simili, Michael Sparks + students)
• Xavier and Matthias for lifecycle management, failure rates etc 

How often?

Monthly or monthly+

Kick-off meeting: organisational meeting, getting people together, explaining what they are going to do

Getting interested people informed 

Put 1-2 pages together and distribute widely (after Simone's review + informing management board)

Need to do this in July (before July 15th)

Where do we report more publicly

OTF for summaries and progress reports

OTF

• Has sessions focused on one topic 
• Ad-hoc meetings in between
• HEPIX for benchmarks, HEP(score)23 per watt
  • They discuss I/O but don't discuss storage as such (or networking). 
  • Storage dominated by embedded carbon (it is never "on or off" unless you really turn off the site).
  • They don't have a physics/CO2 view, they're mostly about different hardware for standard physics workflows
• Need something else for PUE, technical side 
• Needs to go into CRIC 

▾ Action items

DB will line up a few interested people, contact ALICE and CMS, CD will contact ATLAS

CD makes a draft of a document and circulates (can happen early next week)

MS puts together mailing lists (coordinators, common), indico 

Liaising with other projects:

• Question: what is the "official" way to liaise with these external projects?
  • To be discussed with the WLCG management board (Simone) and the management board of the projects.

•We all have informal links, but we need an official contact to their management boards