|25 min||Availability of resources beyond 2020 and how that plays into the XSEDE3 draft transition plan|
Slides presented by Manish Parashar and Sushil Prasad from NSF at PEARC18:
- Slide 11 there is a chart in particular
- Systems retiring near 2020 after a one-year extension
- Remarkable drop by 2020
- John asked if there was a Track 2 solicitation coming up, and Manish said something like, "John, you're connecting the dots" without saying yes or no
- Just got off the phone with Bob Chadduck, program officer, and he has tried to reassure us
- NSF needs to share something with the community, so we can plan for NSF's intentions even if they are unable to deliver
- Bottom of slide indicates ongoing activity
- Manish has taken engaging the community quite seriously and gathering requirements
- A lot of folks providing input, if no plan is shared folks will participate less and continue to wonder
- Shawn - Track 2 to be in by the end of FY19, current BWs system ends before that
- John - there are additional discussions to bridge the gap in time but otherwise XSEDE is illinformed
- Cliff - general discussion topics with NSF, presentation, and general topics beyond 2022
- Take a brand new approach to look at the situation
- So difficult to find out information from NSF
- You do control the connection to the community and what it needs to advance science, XSEDE is in a unique position for that as well as community evolution
- You should be providing that information to NSF, it might influence them
- NSF doesn't do anything as an agency, the people int he community who do things. often times program officers do not recognize that. XSEDE is the agent for NSF. How can we communicate the absolute imperatives for hardware availability and services available
- Hardware systems come and go, always XSEDE adapts and transforms the organization even easier over time
- Communicate that during this time on transition
- Take a bit of control of the situation
- John - issue of on boarding new resources was discussed during the review. Expectation was that XSEDE should predict and develop expertise and Training content to bring the community up-to-speed. Ridiculous that we are to guess. We do provide input to NSF along the lines of what Cliff has suggested but has been rather informal. Are you suggesting something more formal?
- Cliff - door has been opened by NSF, that allows XSEDE to
- Connected to community of scientists and service providers. Could have a conversation with them to determine where their science is going. Take those ideas and impose on them the set of services that you've learned are essential. When new technologies come along, XSEDE will have to do the following based on what you've done in the past. Take those ideas, here are the services that XSEDE will have to have in place for that to come to fruition.
- John - have a side-door way of doing that. Expected to deliver a draft transition plan for XSEDE3 which could potentially put us at a disadvantage
- Cliff - perhaps there would not be a follow-on solicitation, if you take a protective approach there is a risk. You could lay all your cards on the table as a set of what advances science–that is what serves the community best
- Karin - we had a similar situation with Large Scale project for NIH, to provide best information and re-compete against each other. There is also a concern over what budget will be available in the coming months and years.
- John - Bob said something about a stop-and-start government
- Cliff - yes there is uncertainty, but when they have thoughtful documents when it comes time to execute a plan they will be invaluable because they give guidance when decisions need to be made
- Shawn - slide 11 under resources, the so-called innovative systems...there's only a lose connection to what the community needs as a resource. It would be helpful if the solicitations were driven by the science rather than the win, which drives us into proposing technologies that may or may not be useful over the long term. There's a lot of data on systems usage and allocations and could shift the CFP and gets you out of the conundrum of showing your cards
- Cliff - good point, scientists don't always know how to use a new system...it takes x amount of time to adapt and XSEDE has provided those services. It isn't the new hardware. It is how to transition the science to the new technology.
- John - trying to force a bit of transparency on the NSF side. The alignment of simultaneous retirement is worrisome in light of the quarterly lack of resources–2 or 3 times what is available.
- Karin - important observation right now. Consumers are feeling it. Perhaps the XAB could make this more public.
- Laura - state the facts to build a case.
- Cliff - it isn't your problem to take it on because you don't fund it. XSEDE has demonstrated leadership for almost a decade. Continue to demonstrate that leadership. You have the community and experience to do it. Less defensive position and more of a leadership position.
- Shaowen - the future of high-end resources deployed at resources providers. The future of XSEDE is separate. The general state of science from the administration
- John - Resource availability vs. XSEDE...XSEDE can't allocate to machines in 2021, there will only be 2 at that time and begs the question of the existence of XSEDE at all. Calvin Droegemeier is known in the community and could be a good move
- Cliff - if he is appointed, XSEDE should send a letter not about funding but efficiencies and engagement
- Tom - complicated, punt on the training for new resources. If you don't know what you're training for, why train. We had hoped CASC would step into this area. CaRC could. BWs wrote a strong summary that could be used and will send to XAB in confidence. How could we prepare if it takes a couple years to move code. More than just funding for projects and campuses.
- John - provide NSF with input on what is going on in community to support planning process and be ready to respond. Broaden the scope of focus from systems to science
- Tom - letter from XAB could be we hope you see there's a problem the way we do
- Cliff - emphasize taking a leadership and visionary role, because you understand the needs of the community
- Tom - what if XSEDE engaged domains and request white papers on needs and growth
- John - role of the divisions and directorates of NSF, did it 10+ years ago
- Cliff - office of advanced cyberinfrastructure does not understand their role
- Shaowen - how could XAB help? Suggest transparency in planning efforts from OAC
- John - NSF is always welcome to input from the community and the senior point-of-view from the XAB is valuable. It would seem to me that such a letter should offer some ideas about solutions
- Cliff - if this comes from community members it might not be received well
- Karin - whether our advisory board could put forth something, shouldn't we express our ideas
- Cliff - it makes sense, but I understand the internal politics. They could be offended. This is just my personal opinion. If it stays general and doesn't get too specific that might work.
- Laura - science gateways do community planning, perhaps we could tap into that
- Cliff - generalities like GPU usage, those not using parallel processing and how community is using or not using hardware. NSF should know that, because you are in touch with the community.
- John - disruptive technologies that are coming and what new things are on the horizon. There is data that Dave Hart provided that we could use.