© 2007 the university of chicago grid technologies update
TRANSCRIPT
Question
“What would you want the CIO of a research campus, or perhaps this particular group of CIOs, to know or anticipate or think about with regard to grid infrastructures, grid technologies, grid communities, or related?”
Answerers - Grid
Charlie Catlett, ANL CIO, former TeraGrid director
Bob Cowles, SLAC & Open Science GridIan Foster, UChicago/ANL, Computation
Inst.Von Welch, NCSA, GridShib PIChristoph Witzig, SWITCH & EGEE
Answerers - Campus
Jill Gemmill, Clemson, Director of Virtual Organization & Exec Dir for CI Technology Integration
David Lifka, Cornell’s Center for Advanced Computing
Jim Pepin, Clemson central ITJohn Paul Robinson, UAB central ITRenee Schuey, Penn State central IT
What is a grid?
Jim Pepin says
app level software hardware shares colocation of services condo services Grids are evolving towards 'social networks'
as well.
Why Bother with Grids?
Christoph Witzig says
There are advantages for making grids more available to the average scientist (not just HEP): better use of resources access to platforms not available at the local institute lead to better collaboration between institutes and
individual scientists ("the social aspect of science")
Particularly smaller institutions with less resources could profit and pool their resources.
Charlie Catlett says
"I think the biggest issue folks should think about, and try to solve, is identity management."
Campus IdM & Grids
Christoph Witzig puts it thusly
Key is the integration of campus identity management systems with grids and making the access seamless. Rolling-out PKI credentials to everybody is not a viable option.
Grids also need integration with campus networks if they want to expand their user base drastically - so there is also interest on the grid side for integration. It's not a one way road.
Who Will Lead IdM?
But Charlie worries
… we found that this was the thing that tripped us up more than anything else [referring to interop testing among several large grids].
I thought, a year ago, that the campus community (EDUCAUSE/I2) might be solving this but found (at Snowmass in Aug 2006) that this was not the case.
… 15 or 20 campuses in the room saying "gee this is really hard and we have not figured it out".
Common Trust Anchors
Christoph Witzig notes
Proper accreditation of a "automated credential translation" from local credentials to X.509 is very important in order to avoid disconnected "islands of identity".
(Campus?) Attributes
Von Welch says
When someone goes through the whole NSF allocation process [for TeraGrid] they need to be vetted as a member of a valid institution.
Then over the longer term NSF tracks its user community as they move from undergrads, to grad students, to faculty and how their usage progresses.
Shared Resources Require Shared Policy
"The policy / procedure stuff is REALLY tough.
" … the resources contributed to the grid need to be of high quality and that means getting your site central infrastructure people involved - and convincing them they have to do things differently now because they have to have consistent policies (at least for the grid part) as all the other institutions. They are NOT used to operating in that manner."
John Paul Robinson’s principles Respect Autonomy Good Standards Don't Dictate
Implementations. This same goal should be applied to policies.
Charge Back Models Can Be Harmful If your customers are captive (not autonomous)
and your policies (poorly written) prevent them from exploring alternate solutions they will eventually revolt.
Niche?
And he muses
Grids are settling into a small, real but real small, niche.
The last few NSF purchases have been for big honking computers that might plug into Grids but might not. What are the messages here?
The campus more than the national Grids deserve some play.
Readiness Test
David Lifka takes the researcher’s perspective
Are there valuable resources "out there" that make this a reasonable option for my research?
Is the effort required to get my codes working on the grid really worth it?
If I take the time & effort necessary to use the grid will I have priority when I need it? Will there be so many available resources that I'll never have to wait?
Ian Foster says
Universities that want to be competitive in research need to enable their faculty to participate fully in such endeavors.
[referring to increasingly data- and compute-intensive practices for science & scholarship]
Enable How?
On-demand access to substantial amounts of computing and storage (think Amazon EC2 and S3)
Campus grids that provide access to centralized resources (e.g., Purdue) or federate across groups (e.g., UCLA)
Universities host Web pages for their faculty and students--they also need to think about hosting datasets and services
Identity services, Wikis, other collaboration toolsExpertise to help develop and/or install and operate
new applicationsParticipating in Open Science Grid could be a good
start for many campuses
Campus Grid/CI
Jill Gemmill asksWhat CI functions are occurring within schools
and departments, perhaps poorly or in jeopardy, because those functions are not available from you? Would your institution benefit from centralization of these functions?
Are you able to participate as a partner in grant generating activities and receive an allocation of indirect costs?
How can your directory services / authN/Z / provisioning be leveraged for these types of projects? Ditto for storage.
CIO Reflection
Renee Schuey leads us in reflection
What's in it for my institution? What do we have to do to be a user of a grid? A
contributor to a grid? What are the risks of facilitating grid work? What are the benefits to faculty on our campus? How would I find out if there are faculty on our
campus already using grids - and who we might be able to help use them better if we only knew?