2016-06-09 ACME All-Hands V1 Sessions

Please take notes from questions and discussions and post in on this page

Session Time	Duration	Start - End Time	Session / Confluence Page	Plenary Room
Thursday, Jun 9th, 2016
2h		11:05 - 2:20 pm	ACME v1 Session	Session Chair: Shaocheng Xie
	30 min +10 min	11:05 - 11:45 am		Operational Plans on Running v1 Major Experiments (Chris Golaz /coupled)(Presentation + Discussion)
1h 15 min		11:45 - 1:00 pm	Lunch
	30 min + 10 min	1:00 - 1:40 pm		v1 Process for Data Management, infrastructure and Diagnostics (Dean N. Williams (Unlicensed)) (Presentation + Discussion)
	30 min + 10 min	1:40 - 2:20 pm		Overview and Discussion on Papers written/in writing/to be written (Todd Ringler (Unlicensed), Peter Thornton, Phil Rasch (pnl.gov) (Unlicensed)) (Presentation + Discussion)

Change of plans for the first session.

Hight throughput discussion, Requirement for the machine

ne11 2.5 deg watercycle experiment, need input files, compsets, maping files, easily 2 weeks work if it is high priority,
Todd, ok, but no intention to use as a scientific run
Mark, ne11 is the lowest reasonable resolution to be earth-like
land - can you do ne11 case, Phil - they need land domain file and they can do that, who can do the domain file?
Mark, Andy - we also need 10 deg for fast testing to catch the bugs
Mark - if we do one case, we should do two together, 10 deg and 2 deg, extremely low res for SE and very low for scientists

requirement – 40 years per day on the yet to be defined system that we will buy
less then 300 cores, 40 years per day
Chris notes on Edison – problems with the machine unrelated to the code, so there is not good enough support of the machine
Chris would like to automatically kill the hang up job and have another job ready to run, so we do not wait in the queue
We need stable machine environment,
need routinely run tests that the code runs on the LCF
failing exact restart, we did not know about it for some time, comprehensive tests would help with that.
Chris - edison is an interesting machine, it works for some time, then stops and no one knows why
another use is to have the machine for fixing the current failure on the low res coupled model that needs to be fixes urgently
what's the time and effort for the two requirements
Dave - someone will use the super low case and use it for science, we should not release this case code, we could mark it clearly as only for testing