Attending: Debbie Bard, Ken Bloom, Brian Bockelman, Amber Boehnlein, Tim Cartwright, Kaushik De, Rob Gardner, Lisa Gerhardt, Oliver Gutsche, Mike Kirby, Don Krieger, Eric Lancon, Jerome Lauret, Miron Livny, Lauren Michael, Inder Monga, Mark Neubauer, Michael Norman, Benedikt Riedel, Eric Sedore, Horst Severini, Elizabeth Sexton-Kennedy, Frank Wuerthwein
Overview of PATh (Frank)
PATh = "Partnership to Advance Throughput Computing", an NSF Cooperative Agreement that supports the HTCondor team, a "fabric of services" to support the OSG Consortium beyond IRIS-HEP, and a "global infrastructure laboratory" (whose purpose is the ingest of ideas and software into PATh).
NSF has a vision of a computational infrastructure that includes leadership class systems, innovative capacity capabilities, exploratory pilots and prototypes, and federated resources. PATh is in the last of these categories, and the only one in that project right now. Cloud services and campus CI span all four areas, so PATh touches on cloud too. But it is most strongly aligned with the CC* program, such that OSG is listed in the CC* solicitation. Whenever there is a new CC* proposal deadline, PATh is there to talk to people who are new to the program and help them write the strongest possible proposals. When awards are made, PATh helps make the 20% of a campus resource available to the community. This makes PATh a bit like XSEDE in helping to provide community access to a resource.
Miron notes that having a stream of quality proposals into CC* is part of the strategy of PATh; there is value in having funding for resources that can then be federated, and Miron sees them as symbiotic. Inder asks why there is not the same alignment with CloudBank; Frank replies that it is mostly because they haven't figured out how to do so. CloudBank appears to be mostly about the financial arrangements for the cloud; there is more to figure out here. Mike notes that CloudBank is happy to find ways to work with OSG and interested in forward-looking experiments.
PATh is also expected to coordinate with other coordination services, including e.g. XSEDE and XDMoD and their successors. It is anticipated that there will be more coordination services in the future; PATh is just an early arriver in this paradigm. The scope of PATh might change as NSF sets up more of these services.
OSG-LHC within IRIS-HEP covers things that are relevant to the LHC that both US CMS and US ATLAS agree on. This includes the OSG software stack and support, support to US LHC T1 and T2 (whereas T3 is part of the CC* community and thus under PATh for support), operational security, operations of accounting et al., and network performance measurements. It is ~5.2 FTE over 13 people. These things are thus not covered by PATh.
PATh looks after a variety of services instantiated via software. These include submission endpoints, resource pools (including but not limited to the Open Science Pool), CEs ("Computing Entry points", rather than Compute Elements!), and resource provisioning and annex components (e.g. glideinWMS). There is also all of the data and software management: data endpoints, XRootD, data caches, Rucio as a service for the community, and more. This all supports the data federation and more. EGI was recently funded for similar tasks in their region. Note that OSG is already operating Rucio for some VOs; much discussion about that. Mike notes that this could be a good joint pilot project with CloudBank.
There are also services provided by people, such as facilitator activities. New to this area is virtual office hours and long-term follow-up as part of ongoing guidance to users.
There are a total of 20.36 FTE on PATh. There are four big areas: Software Development (8.40 FTE), Global Infrastructure Laboratory (0.93 FTE), FoCaS (9.34 FTE), and Community Empowerment (0.55 FTE). Plus a project office, 1.14 FTE. The Software Development area is responsible for HTCondor. This leaves about 12 FTE for OSG activities. Add to that 5 FTE in IRIS-HEP for about 17 FTE, compared to about 23 FTE in the "old" OSG. Overall it is less effort than at the start of the five-year project that was funded as OSG. The biggest change is the operations part; there is no GOC anymore. But there are also evolutions, and OSG has gotten better at what it does so there is space to do new things. For instance, when N5Y started, there was no good story around data (or, the IRODS effort collapsed). Miron notes that the original OSG award was very LHC/LIGO-centric; NSF was very concerned that LHC would succeed. The N5Y award was more of a mixed bag. But today the funding is from OAC, with the understanding that LHC needs are covered by IRIS-HEP, so a very different community is served now. And today's team is much more efficient.
What's the interaction between the Council and PATh? No different with the past. It interacts with the OSG ET. The PATh and IRIS-HEP PIs are on the Council. Any other project that will provide effort to the consortium would presumably become a member of the Council. PATh will follow the broad direction and vision that we have collectively had. While we claim to NSF that we've changed our vision, we haven't really; we're just finally achieving things that we've always wanted. It only took about 16 years to get here. We had to make LHC work to be allowed to pursue this vision. Miron says that there is a relationship between the Council and Frank, who is the elected ED. PATh is not OSG, which has the Council and ED and effort that comes from different places. PATh has a PI and co-PIs who are responsible for the project. It views the Council as an advisory body. PATh is accountable to OAC. In the past, Miron as a PI was responsible to MPS, a different entity. In his opinion, the Council makeup more reflects the history of OSG than the future.
Rob says that in the past we've had a global picture of how all the projects add up into OSG. Is there a role for the ET and the Council to incubate the next partnership project that might fit into the OAC vision? Yes; Frank would like to see OSG gain more traction in agencies other than NSF (e.g. DOE but not just that), to get another IRIS-HEP-like thing focused around another agency's mission. DHTC could be relevant in many other areas. Some effort would be required to make these matches. Miron says PATh was an unsolicited proposal, and it was a lot of work to position it in the right way. Dedicated PI effort would be needed to bring in some of these other science areas. Miron is proud of the effort that has been brought in, and would appreciate some appreciation from the Council. He would like to see the Council provide more support for the next proposal five years from now.
Campuses on the Council (Miron, Eric)
Miron and Eric discussed and decided that especially since Eric has his own CC* award, he can be a powerful contact point with existing and newcomer CC* campuses, and more credible here than Miron. Eric says that there is a lot of good work going on, and learning about what is going on and figuring out how to represent OSG to his peers is where to start. This will be a long-haul effort; campuses would be picked up at a slow but steady pace in the utilization and providing of resources.
Miron would like a focus on the number of users reached rather than SUs provided; Mike agrees that this is the right sort of question for the "long tail". Campuses can help us understand who these users are. Further discussion on how we measure and track impact. The Council could be helpful here. OSG accounting can be helpful in showing the value we provide. We will try to explore this for a future Council meeting. Debbie Bard will organize an effort with several others on this. More can be done to gather information on publications and other metrics of impact via surveys. We could also find ways to highlight student contributions.
We will need to elect a deputy Council chair in Spring 2021. This person would become Council chair in Spring 2022 and serve for two years. Ken and Rob will appoint a search committee; volunteers should contact them.
Miron reminds everyone that Internet2 is having an event today about the impact of NSFnet, and at the end the AD of CISE will be talking about the Science Cloud, which could be informative.
Horst is not convinced that the spring AHM can be hosted at OU, and that most likely it will be virtual. Among other things, there will not be a spring break week at OU. Rob agrees with this. The projected date has not yet been set, and it needs to be. Frank will take this as an action item.