Skip to end of metadata
Go to start of metadata

When: Thursday, February 25 from noon - 1pm
Where: 200C Warren Hall, 2195 Hearst St (see building access instructions on parent page).
Event format: The reading group is a brown bag lunch (bring your own) with a short <20 min talk followed by ~40 min group discussion.

Presenter: Jason Christopher, Research IT
Facilitator: TBA

====

Researchers need easy access to analytic computing environments that are designed to fit their needs. BRC's new Analytics Environments On-Demand (ÆoD) service, currently under development, is intended for researchers who need to run analytics software packages (such as Stata, SPSS, R Studio, etc.) on a platform that is scaled up from a standard laptop or workstation, in an environment that is readily accessible and secure. This talk will explore approaches to providing analytics environments on demand for researchers across the university.

Please review the following prior to our 2/25 meeting:

 

 

Presenting: Jason Christopher, Research IT

Attending:

Aaron Culich, Research IT
Aron Roberts, Research IT
Barbaa Gilson, IST-Administrative Applications
Catherine Carson, Data Science Initiative
Chris Hoffman, Research IT
Chris Paciorek, Statistics & BRC
David Greenbaum, Research IT
David Fulmer, School of Social Welfare
Erik Mitchell, Library
Jason Christopher, Research IT
John Lowe, Research IT
Jon Stiles, D-Lab
Kelly Rowland, Nuclear Engineering & BRC
Kim Carl, VCRO
Norm Cheng, TPO
Quinn Dombrowski, Research IT
Patrick Schmitz, Research IT
Rick Jaffe, Research IT
Ronald Sprouse, Linguistics
Ryan Lovett, Statistics
Scott Peterson, Doe Library
Steven Carrier, School of Education
Steve Masover, Research IT

 

Notes

See Jason's slide deck.


Patrick: Curious to see how AEoD will intersect with tendency to provide portals as a means of delivering functionality without requiring researchers to master the ins and outs of a sometimes-complex new environment (such as HPC or national infrastructure or AWS).

Jason: Underlying techologies (Citrix, Windows Terminal Server) are geared to enterprise needs (everybody does the same thing) rather than research (many project have particular needs, edge use cases). Variety of storage options wanted by researchers are, similarly, not supported or poorly supported by these technologies.

David Greenbaum: Related initiatives in this space -- Erik/Library Citrix effort; Library/ETS/Student Affairs re: student computing support. D-Lab? What is the campus landscape in this space, and where should investment come from.
Jason: Looking at and experimenting with service as offered by Library for students to examine how it does & doesn't intersect with researcher use cases.
Aaron: E.g., researcher use of specialized modules -- extra add-ons; also demands on the system resources. Delineate differences.
Erik: Our service is highly standardized, that was the pitch we made to the Citrix team; we're willing to roll with that given who we serve (students) -- different from use cases that Jason just outlined. David Fulmer is experimenting with SPSS - Data Center storage connection, for secure computing with this service.
David Fulmer: Data over network caused large-dataset jobs to fail; looking to see if we can solve that by locating data in same data center as resources running SPSS atop Citrix.
David Fulmer: Very strong interest among PhD students in Social Welfare in having this functionality available remotely in order to free their laptops and other local resources for other tasks.
Erik: Kind of shocking how many groups we had to work with to make this experiment fly. Speaks to the frustration of working at Berkeley. How do we keep all the pieces in synch over time: durability. "Too shoestringy" for an institution of this caliber.
David Greenbaum: is that complexity a function of no one stepping forward to be a service owner.
Erik: Would be great if Research IT became service owner. We thought it would be less difficult than it turned out to be, esp. in light of licensing (Stata, SPSS, Windows) constraints.
Erik: Unified Computing Group answering a different question than AEoD. Simplifying at the client. Providing a baseline of apps/functionality across the whole student population.
David Fulmer: Yes, ditto for Social Welfare. Would like to offer incoming students a Chromebook at $200 price point through which students can access all the apps they need to do their work here.
Jon Stiles: Have been focused on instructional rather than research context at D-Lab.
Chris Hoffman: If other institutions -- UCSD, Pittsburgh (Bridges) -- are running services, should we get someone else to run it and purchase from them?
Aaron: Bridges & Jetstream are options in national infrastructure, which is set up to expect a campus infrastructure supporting local needs. We're looking at how to leverage new models that are emerging. Cloud WG meeting in April will look at the new XSEDE offerings.

Rick: Tension between provider wish to offer a standard offering, and researcher need for customization. Is there a sweet spot that will actually satisfy needs?
David Greenbaum: A "business to consumer" model; or a "business to business" (where the receiving business is school of social welfare, school of education, etc.) where the receiving org handles the "last mile" direct user support.
Jason: B2B is probably the model we can support. High-touch, highly-variable service to end users is almost certainly too expensive for BRC to support.
David Fulmer: End users (researchers) are v. rigid about service offerings -- if it doesn't work first time, forget it, they're gone. For longtime research projects, high resistance to making changes years or decades into a research program.

David Fulmer: Worth mentioning that when Jason mentioned faculty using services at an institution where they were located previously, 3 faculty members came immediately to mind. Their prior institutions offer services that Berkeley doesn't. End of story (until we offer something more / different).

Chris P: Timeframe for rollout?
Jason: preliminary testing in June
David G: Service definition and design inquiry. Questions about cost, where's right place to locate this service, comparison to what else is offered on campus. So there will be some close examination about where the right investment should and can come from before we are able to offer a production service.

 

  • No labels