[an error occurred while processing this directive]
[an error occurred while processing this directive]
Minutes of the CSM SSC Meeting Monday, 21 June 1999 The Village at Breckenridge
Minutes of the CSM SSC Meeting
Monday, 21 June 1999
The Village at Breckenridge
Attendees: Blackmon (Chair), Solomon, Shukla, Randall, Dickinson, Bamzai, Kiehl Boville, Bryan, Gent, and Harper
1. Review of Action Items
a. Blackmon reported that the Senate subcommittee recommended not to fund ACPI. DOE will continue its efforts to reverse that decision. Blackmon will meet with Dannevik and Malone regarding the NCAR/LANL/LLNL consortium on Wednesday, 23 June 1999.
b. Gent reported that the CCM3 with prognostic water, x2' ocean model, land model, and VP sea ice model with river runoff switched on is running on the SGI. He also reported that the throughput on the SGI is not as high as expected.
c. Blackmon reported that initial discussions have taken place with Unidata to compress and parallelize netCDF. Due to the recent death of a key employee in Unidata, no commitments were made. Further discussions will be held in the future.
d. A list of CSM-related current awards at NSF was distributed to the SSC.
e. The SSC said that the "draft" CSM Data Policy Statement was a good start and that CSM PI should be defined more specifically as "the person who designs and runs the experiment." Blackmon requested any other comments be sent to him via e-mail.
f. Blackmon reported that the revised CSM Advisory Board (CAB) Terms of Reference were accepted and that the CAB would give advice to the CSM SSC.
2. Summary of CAB Meeting. Blackmon reported that Suki Manabe resigned from the CAB and will be replaced by Tony Hollingsworth, ECMWF. The CAB also added Max Suarez, NASA/GSFC, as a member. The overall CAB attitude toward the CSM project is that it is doing well. Blackmon also reported that the CSM Workshop in 2000 would be the last full week of June (27-29 June), and then in 2001 the CSM Workshop would move back to the 3rd week of June. The CAB advised the SSC that specific details of who appoints working group co-chairs and who writes the CSL proposal, etc., should be written in the CSM Plan.
3. Summary of Computing Issues. The type of computers NCAR will purchase has been very confusing over the last months. The plan at this time is to buy a cluster of IBMs with 128 nodes with two processors per node. In October, 1999, NSF will hopefully fund a $5.5 million expansion to bring the NCAR computing to 256 nodes with two processors per node. Then in April, 2000, NCAR hopes to swap the two processors per node for upgraded nodes, each with 4 higher performance processors. Nine months later another upgrade is expected with 8 or 16 processors per node. A small test cluster of Compaq computers is also expected. The Cray C-90 will be decommissioned as of 30 September 1999, the Js will also be decommissioned, and the SGI Origin will be supported at least one year and maybe longer. The CSL will include the SGI Origin, two J se's, one-half the IBM cluster, and in October will be augmented by a small test cluster of Compaqs.
The SSC discussed the impacts of the NCAR computing system on CSM. The SSC decided that we would continue to run the CSM 1.0 for a short time on the Origin and then let it die. Our greatest efforts will be on running the CSM component models and coupled model on the IBM cluster. The Community Climate Model (CCM) will be recoded for the IBM architecture, the POP model is already being worked on to run on the IBM at LLNL, the sea-ice model is already running on the IBM, and the land model is no problem because it is already parallelized. The SSC also discussed writing a letter to SCD about keeping the C90, but there was no consensus to do so. Some SSC member's thought the $1 million maintenance cost of the C90 would be better spent on upgrading computing.
Boville will lead the CSL proposal writing for the CSM project. The SSC requested that Fein (via conference call) discuss with SCD a postponement of the CSL proposal deadline until December with allocations beginning 1 January 2000. Because a unit of measure for allocations has not been decided for the IBMs, it would be difficult to propose experiments. When the CSL proposal deadline is decided, all CSM working groups will be asked to submit 1-1/2 pages via e-mail to Boville that includes their science plan, details of experiments they want to run, and costs of each experiment. A CSM SSC meeting will be planned to discuss the CSL proposal for CSM. The SSC will also request a waiver of the length of the CSM proposal via Fein to the CSL Allocation Panel and SCD.
4. Discussion of CSM Plan. Blackmon reported that a revised CSM Management Plan had been drafted to include SSC oversight of the CSM working groups, appointment and replacement of CSM working group co-chairs, and other suggestions from the CAB. Blackmon requested advice from the SSC to define what the lack of funding impact was on CSM and also what could be done with additional funding. It was agreed that additional funding could be used to hire support staff at NCAR to recode and improve the CSM. Blackmon recommended that he and Randall receive updated working group plans by 1 September 1999, so that they could begin to write a cohesive science plan. Blackmon will request updated working group plans at the Thursday plenary session at the CSM Workshop. Blackmon will visit each working group meeting and request that working group meetings be held in the winter, or at the latest in March, so that there is time for experiments to be completed and results to be shown at the CSM Workshop in June. The goal is to have a final plan for the CAB to review in December/January timeframe. Comments will be needed from the SSC members rapidly on the draft CSM Plan.
Boville suggested a new CSM Software Engineering Working Group be started that would include all CSM component model programmers to discuss and recommend ways for the coupled system to be assembled, sequential execution, multiple processors and parallelization, flux coupler issues, etc. Kiehl said it was absolutely necessary to start this working group because there is little communication now between model component programmers. The SSC recommended a fall meeting at NCAR with invited guests and NCAR programmers that hopefully would be sponsored by Serafin, and they recommended appointing two NCAR people to coordinate the meeting and agenda. A charge should be written and given to the appointed coordinators by the SSC. A request for funding of this initial working group meeting will be sent to Serafin.
5. Advice on CSM Funding Priorities for NSF CMAP. The SSC received a proposal from Sumant Nigam to create a community tool for diagnostics. Randall said diagnostics that would be done at NCAR and universities would be a step forward. Shukla and Dickinson did not think that diagnostics should be singled out for funding. Solomon suggested scaling back Nigam's proposal to 2 groups instead of 5. The SSC recommended that Fein discuss the diagnostics proposal at a smaller scale with Nigam.
6. CSM Working Group Co-Chairs. The leadership of the CSM working groups was discussed.
7. Component Model Processing Suite (CMPS). Sylvia Murphy gave a presentation on the new CMPS that will provide processing tools for all CSM component models. Shea and Alpert also attended to answer any questions on the future plans of CMPS. Blackmon asked for any comments on the processing tools and future processing tools be sent to him.
8. Discussion of the Seasonal-to-Interannual Work with CSM. Zebiak, Ji, Anderson, and Tribbia joined the SSC meeting to discuss SI issues. Tribbia reported that there were reasons for SI collaborations to be and not to be included in the CSM project. SI and CSM have overlapping interests in ENSO/teleconnections and computing; the ability to economize resources (manpower and computing); and synergism in prediction, diagnosis, and development. Tribbia also noted that SI and CSM goals are dissimilar (prediction vs. simulation); they have competing needs (accuracy for SI vs. centennial feedbacks); responsibility and ownership is important in each group; and there could be dilution of efforts. Zebiak stated that IRI's mission is to advance the state of SI climate predictions. Anderson said MOM4 might be ready in nine months (February/March, 2000). Ji said NCEP has IBMs now and that use of the IBMs was opened internally two weeks ago. MOM has data assimilation already and POP does not. Ji stated that a standard flux coupler between IRI, NCEP, GFDL, and NCAR would be beneficial. Zebiak stated that IRI wants to work with the CSM and hopes the SSC will support that collaboration. It was stated that CSM is very important in the "framework" issue.
9. Miscellaneous. Blackmon reported that copies of the Polar Climate, Ocean Model, Paleoclimate Model, Natural Variability, and Seasonal-to-Interannual Working Group Reports were included under Tab N.
The SSC agreed to have their next meeting in November to discuss the model conversion issues, the schedule for CSM 2.0, the updated CSM Plan, and the CSL proposal. Harper will request SSC member's schedules in November and early December so that this meeting can be scheduled.