Memo: The NASA ESE Computing Requirements Workshop
From: Karl Lalonde

I attended a workshop in Washington, D.C. hosted by the NASA Earth Science Enterprise group from April 30th through May 1st. The purpose of the workshop was to address computing requirements and ascertain data storage needs for this branch through the year 2010. Three separate groups of researchers were in attendance: Climate, Weather, and Solid Earth research. Roughly 150 people participated, many from NASA headquarters, JPL, and Goddard, with a scattering of representatives from the academic realm. There were no preconceived notions as to the outcome of the workshop. NASA ESE wanted specific recommendations and direction from the participants.

The original workshop structure was to split into three groups representing Climate, Weather, and Solid Earth on the first day. These subgroups would isolate specific requirements for their respective areas, and then report at the end of the day to the entire group.  The second day would consist of splitting into three cross-cutting groups of equal size developing specific requirements to offer to NASA ESE. Didn’t quite work out this way.

The weather and climate groups were fairly large, so I decided to participate with the Solid Earth research section. This involves things like volcano and earthquake prediction, mantle modeling, geomagnetic flux monitoring and modeling, and crustal plate interactions. Many of the models are apparently new and just coming online. Some of these models have between 1000 and 3000 variables and the model domain is the entire planet, so computing power is definitely an issue. This particular group is also going to be on the receiving end of data from several new satellites as well. My background being somewhat geologic and this section being the least represented, I thought my ability to contribute (if necessary) was optimal in the Solid Earth section.

The first day was an exercise in self organization. There were subgroup chairs, but nobody in any of the subgroups really knew how to start, and there was intentionally no direction offered by the workshop hosts. Our approach ended up being to figure out exactly how much data was going to be generated by the models and the new satellites and then reverse engineer storage and processing requirements, and project those requirements eight years into the future. All of this information was put into a Powerpoint presentation (one of two which are included with this report) and offered to the entire workshop at days end. (Web editor's note: The two PowerPoint presentations are on file at SDSM&T). Interestingly, data generation and storage was the primary focus of the solid earth subgroup, as well as the other two subgroups. I opted to point out that data understanding in the forms of data analysis and visualization would be beneficial and also computationally intensive. These ideas were accepted and included in the Solid Earth offerings.

The NASA ESE heads met after the first day, and decided that though many abstract problems have been resolved, the goal of specific recommendations had not been met. We stayed in our original groups and hammered out details that were more in line with the workshop goals. These recommendations were presented at the end of the second day to the entire workshop and are included in the second Powerpoint. SD Tech’s contribution was making sure that data understanding/visualization would be included in the recommendations to NASA ESE. 

Additional observations include:

1)                 Some real concerns were voiced over the realities behind chip technologies in general. Memory really hasn’t gotten faster and chip innovation, short of simply enhancing speed as a marketing technique, hasn’t really evolved. These are issues when trying to do high performance computing with commodity clusters. 

2)                 Methods of getting data from distributed multiple sources in real time for model infusion and natural hazard prediction need to exist.  

3)                 Model development needs to be treated as a planetary science problem. Develop models that can run on Mars and Venus, as well as the Earth. 

Finally, I appreciate being able to participate in this workshop. I think we contributed something to the end result.


Back to "Trips and Collaboration" SD Personnel and NASA Centers / NASA-related activities website