Welcome and discussion of the process NSF plans to use during the next round of re-competition of XSEDE
John Towns
Lessons Learned from TeraGrid seem to have been lost by NSF
NSF process constructed that the awards being made are smaller in size
How do we advocate for what we've got without sounding self-serving?
Dave Lifka - difficult to justify a large project to congress
Albert Lazzarini - could it be that NSF is reflecting what they are hearing from the community?
Ralph - if you want to recompete XSEDE, it would be difficult for another team to put something together. Bank being too big to fail; therefore break it up so the competition is manageble. There are very severe consequences. What was missing in TeraGrid was central communication.
Cliff - don't buy the argument for large award. Therefore, what is the best approach? Attitude you should have: we are in a good position to provide computational capability to the community, we think there are multiple approaches. Present the options for going forward in a small report, stating that the current situation is a good situation. NSF doesn't know what it wants, but it knows its pressure. Providing a report gives thoughtful community input. Acting as an agent for NSF.
Karin - responsibillty as the XAB to perhaps provide this report.
John - TeraGrid Futures had community input, NSF largely didnt' listen to the recommendations
Cliff - NSF is reorganizing again.
John - Rudi's term will be up in June an do not yet know who the new program officer will be
Nancy - could we pitch it as we are not the only institutions who can, but we could present it as this is the best structure to move forward
Cliff - help NSF achieve its goals through a set of options
Shaowen - time is ripe for a more defined process and structure for digital infrastructure; for example, NCAR getting sustained support. Solid community structure perceived as objective and representative. While the XSEDE work is phenomenal, it is to a degree ad hoc–how can we establish a structure for a long-term vision for the community.
John - NCAR is an FFRDC the funding threshold for MREFC has changed and is now at $70M. Not advocating that XSEDE become an MREFC
Cliff - value in FFRDC, but it is a cultural issue. We could achieve much more with the building infrastructures than with individual grants.
Dan - matching risk profile to budget and managing budget, kicks in at $50M
Ralph - XAB approach the ACI board to think of alternatives
Cliff - we would have to have OCI bring it to NSF's attention, not sure OCI would do it
Albert - program advisory committee selects and advises and has no relationship with NSF. XSEDE management could formally ask for XAB feedback and share feedback with NSF
John - it has to fall back to a process NSF is comfortable with. Recommendations from us and the community to NSF, but XSEDE wouldn't be driving it.
Cliff - ask NSF their vision 10 years from now as a foundation
Shaowen - long-term vision answer might be to ask community input
Karin - workshop for community to come up with vision
Cliff - we want to figure out how to engage them. Thom Dunning working to respond to larger reports, there might be kernels of information that feeds into the recommendations.
John - haven't been engaged in that process, concnered they don't want XSEDE involved
Shawn - between the broader community and the XSEDE community is the user community–how do we engage that community?
John - planning grant exercise
Shawn - plenary town hall session at SC17 asking for feedback
John - have done similar exercises. Challenge, SC is appropriate, but XSEDE is broader than SC.
Shawn - do how to you reach 80% of your users?
John - user surveys, focus group discussions. Holding several town hall meetings over the course of a year or more.
Cliff - focus on the science. OCI will be receptive to a long-term strategy gathered at Science conferences as a way to accelerate discovery in particular domains
Ralph - when the original XSEDE grant was made, it was stated that it could be renewed without a public competition; that is what they removed this time
Shawn - how could the characteristics of what XSEDE provides vs. what XSEDE should provide conflate the issue
Ralph - most would say we have a lack of resources
Tom - could re-compete subsets
Program office
Operations
XRAS
CEE
ECSS
XCI
John - requirements for success. Because we don't have independent PIs, they don't behave in the best interest of them–they work collectively toward what is best for the program as a whole. The reach of XSEDE should be broader, NSF felt XSEDE is the core. Concern is fracturing the core.
Ralph - the L2 areas were decided by NSF for XSEDE2. Different than XSEDE1
John - XSEDE is flexible to change, but an award structure creates a certain set of permanence. Have to deliberatly change through a process. We will be coming back to this issue. Clear that more concrete plans are needed for NSF. Going in a direction that XSEDE agrees with and figuring out how to address that.
Cliff - success you've had should allow playing a role in defining the future going forward