DSPy a system for AI to Write Prompts and Do Fine Tuning
ARCC National Perspective Panel: XSEDE (Towns)
1. March 17, 2014
Research Computing on Campuses:
National Perspectives Panel
John Towns
PI and Project Director, XSEDE
Director, Collaborative Cyberinfrastructure Programs, NCSA
jtowns@ncsa.illinois.edu
3. Who is this guy?
• John Towns
– PI and Project Director of XSEDE
– Director, Collaborative eScience Program Office at NCSA
– Co-Founder, Illinois Campus Cluster Program
• Background
– just a mid-western boy from Missouri
• I don’t see a lot of folks from Missouri in the circles I am in
– failed physicist
• general relativity
– failed computational scientist
• numerical simulation of black hole spacetimes
– failed networked applications guy
• NLANR-DAST: National Laboratory for Applied Network Research-
Distributed Applications Support Team
– budding social engineer
• Wearing my XSEDE hat for this discussion
3
4. XSEDE – accelerating scientific discovery
• XSEDE’s Vision:
a world of digitally enabled researchers, engineers, and
scholars participating in multidisciplinary
collaborations to tackle society’s grand challenges
• XSEDE’s Mission:
to substantially enhance the productivity of a growing
community of researchers, engineers, and scholars
through access to advanced digital services that
support open research; and to coordinate and add
significant value to the leading cyberinfrastructure
resources funded by the NSF and other agencies
4
5. XSEDE’s Strategic Goals
• Deepen and extend the use of the advanced digital research
services ecosystem
– deepen use by existing researchers, engineers, and scholars
– extend use to new communities
– prepare the current and next generation via education, training, and
outreach
– raise the general awareness of the value of advanced digital services
• Advance the advanced digital research services ecosystem
– create an open and evolving e-infrastructure
– enhance the array of technical expertise and support services offered
• Sustain the advanced digital research services ecosystem
– assure and maintain a reliable and secure infrastructure
– provide excellent user support services
– operate an effective and innovative virtual organization
5
6. What is XSEDE?
• An ecosystem of advanced digital services accelerating
scientific discovery
– support a growing portfolio of resources and services
• advanced computing, high-end visualization, data analysis, and
other resources and services
• interoperability with other infrastructures
• A virtual organization providing
– dynamic distributed infrastructure
– support services, and technical expertise to enable
researchers engineers and scholars
• addressing the most important and challenging problems facing
the nation and world
• A project funded by the National Science Foundation
6
7. Discuss a particularly challenging/thorny issue… (campus cluster relevant)
Campus Bridging
• How do we facilitate campus bridging?
– integrating campus researchers and resources into the national
cyberinfrastructure ecosystem
• Multi-pronged approach
– Campus Champions
• local source of knowledge about HPC/HTC and other digital services,
opportunities and resources
• empowers campus researchers, educators, and students to advance
scientific discovery
– Campus Bridging
• lower the barriers to utilization of XSEDE resources by researchers and
to disseminate the best practices of XSEDE resources to campus IT
staff
– Service Providers (SP) Forum
• a forum for open and inclusive discussion of topics of interest to the
SP community
• a formal communication channel between the Forum members and
XSEDE
7
8. Describe your top unsolved challenge…
Articulating the value of XSEDE to stakeholders
• Stakeholders
– funding agency: NSF
– community we support: researchers, scholars and engineers
– our partners: Service Providers, other
infrastructure providers, collaborators in
technology development,…
– administration at our respective
institutions
– ourselves: XSEDE staff
• Requires clear metrics of success
– How do you clearly and succinctly relate
science impact when you don’t do any science?
– How do you clearly and succinctly relate societal impact to “Joe
Six-Pack?”
8
9. Describe your top funding problem…
Sustainability of Funding
• XSEDE in year 3 of 5 year project
– Can we secure an additional 5 years of funding?
• How do we sustain after 10 years?
– Can we avoid disruptive nature
of re-compete while still
deriving best value?
– Will NSF even continue to
fund such an activity?
• Is NSF definition of “sustainable”
that they no longer fund it?
9
10. How would you spend $1B (US/CAD) to dramatically improve our national CI?
Multiple Efforts! (some of what I would do…)
• Co-invest with campuses to develop campus CI: $200M
– ??% match from campuses; must integrate with national CI
– 2-3 rounds of awards running 3-5 years
• Establish a national data infrastructure: $250M
– needs to compliment the national computing infrastructure
– 2-3 rounds of awards running 3-5 years
• Workforce development: $40M
– expand current and grow new programs to develop CI-savvy
• researchers: grad student, postdocs, research scientists, faculty, …
• technical staff: systems admins, network engineers, applications support staff, …
– 2-3 rounds of awards running 3-5 years
• Develop new leadership in our community: $10M
– establish programs to develop leadership
• center directors, facility mangers, large scale project leads do not have these
– 2-3 rounds of awards running 3-5 years
• Sustain this! $500M
– investment fund to generate ongoing funds for sustaining/expanding these
efforts
10