Date: 6-16-2015, 0800-0915 CT (GEC23, University of Illinois, Champaign-Urbana)

Participants: Nick Bastin, Max Ott, Ivan Seskar, Mike Zink, Marshall Brinn, Mark Berman, Chip Elliott, Aaron Helsinger.

Agenda

The agenda of the GEC23 meeting of the GENI Architect team was to focus on the future of GENI from an organization as well architectural point of view. What processes do we need to make and disseminate decisions about the GENI architecture? What themes and functional capabilities should we stress? How should GENI position itself with respect to other cyberinfrastructure initiatives in the coming years?

Summary

Several major themes emerged from the meeting.

Looking at GENI going forward, there seem to be several different goals:

1. A stable GENI with relatively static capability that can be used for education and research
2. An experimental GENI with new kinds of resources (new rack/switch times and entirely new kinds of resources like GPUs and FPGAs, etc.).
3. A federated GENI that grows by linking and interoperating with other cyberinfrastructure testbeds and clouds and helps develop new standards and technologies for federation, SDX, policy, identity.

These may be conflicting goals and as GENI grows, it may evolve along different paths simultaneously. “Evolve and dissolve”: Both continue to grow and become part of what CI is prevalent in the future.

We discussed the challenges of entropy of software and of documentation, tutorials, manuals, schemas. The obsolescence of hardware is a similar concern. One charge of an ongoing architecture team will be to push these things to be kept up with changing times and requirements or brought off line.

Several of us stressed the importance of adding new innovative, even “bleeding edge” resources to GENI, that there are diminishing returns for a larger GENI consisting of the current set of resources and that different research will be driven by providing access to new and rare kinds of resources (much as OF HW switches were when the first GENI racks were being specified).

We talked about the challenges, past and future, of getting campus buy-in for GENI and how that might be alleviated moving forward. Clearly campus CIO buy-in is critical (top-down) as is researcher demand (bottom-up) and the architecture team may play a role in mediating a conversation between these entities.

We discussed the future of the GENI architecture team as well. The goal of the coming two years is to ramp down GPO leadership and transition it over to the community, and the GENI architecture team is no exception. We thought that there should be a body to make policy decisions about upgrades, review API changes and federation decisions, decisions about adding new resource types. Perhaps, though this body should extend beyond the scope of GENI and operate more at a level of “NSF CI architects”.