A Workshop on Effective Approaches to

Campus Research Computing Cyberinfrastructure

April 25-27, 2006
Crystal Gateway Marriott Hotel
1700 Jefferson Davis Highway
Arlington, VA 22202

Home || Workshop Proposal || Agenda/Presentations || Roster || Final Report

Workshop Flywheel: Steve Olshansky, Internet2

Schedule:

• Tuesday April 25, 2006, 3:00 PM - 5:30 PM. (for CIOs and CTOs)
• Wednesday April 26, 1:00 - 6:00 PM (for CTOs)
• Dinner Wednesday evening 7:00 - 10:00 PM (for CTOs)
• Thursday April 27, 8:00 AM - Noon (for CTOs)

Summary:
Historically, campuses have provided some measure of research computing capacity to scientists and engineers. With the advent of supercomputing centers and departmental facilities, campuses began to reduce their involvement in provisioning computing cycles to research communities. Now, a richer understanding of the realities of cyberinfrastructure reinforces the value in having greater enterprise-level research computing capabilities. Moreover, if those capabilities were deployed in a consistent manner, then efficiencies, greater user effectiveness, and the possibility for inter-institutional leverage could occur.

This workshop, aimed primarily at major campus CTOs but with relevance to CIOs, will begin a community process for greater common understanding around the technology options and, to a lesser degree, the political and financial issues for campus research computing cyberinfrastructure. The workshop will be centered around coordinated presentations by several universities that are doing leading-edge deployments of research computing capabilities. Campuses will be selected to present an architecturally diverse set of approaches, including campus-wide Grids, clusters (e.g. Condor pools or Beowulf clusters), condo’s, etc. In order for the workshop to have impact beyond the immediate participants, it is important that it provide outputs beyond a website of powerpoint presentations. To that end, we intend to deliver a whitepaper detailing the approaches and the comparative analyses that emerge from the workshop discussions.

We also intend to create a list of issues for further consideration by NSF in its goal of promoting campus research cyberinfrastructure. Lastly, we intend to take advantage of the many outreach opportunities that Internet2 offers for dissemination of the white paper and workshop results.

• Kevin Morooney, Pennsylvania State University (Chair)
• Patrick Dreher, Massachusetts Institute of Technology
• James Jokl, University of Virginia
• Kenneth J. Klingenstein, Internet2
• Matthew R. Link, Indiana University
• Stephen Moore, Georgetown University
• James M. Pepin, University of Southern California
• Gordon K. Springer, University of Missouri-Columbia
• William Wichser, Princeton University

Workshop Goals and Outcomes:
The workshop is divided into two sections. The first, a higher-level part of the program will identify basic technical options and discuss motivations and issues for campuses to consider as they move forward. It will also map out and develop the sets of issues that will need to be worked in creating a true multi-campus sharing, from technical aspects to policy and financials factors. The bulk of the workshop will consist of a set of coordinated presentations from major research universities that are or are implementing campus research computing cyberinfrastructure.

The presentations will focus on the following sets of issues:

• Technical architectures: There are several major alternatives for research computing cyberinfrastructure, including Condor pools, campus Grids, Beowulf clusters, BlueGene class systems, and condos. Campuses will be selected to present that cover this full spectrum of implementation choices. Discussion topics will include cost/performance issues, physical facilities issues, fit of architectures to application requirements, issues in linking resources across multiple campuses, etc.
  
  
• Associated data storage issues: Many high-end computation problems require large amounts of data and perhaps intermediate storage. Locating the data, striping and other appropriate staging approaches can make considerable difference in performance, etc. and the campuses will be asked to present on the options they chosen, rationale, and levels of satisfaction with the option selected. Additionally, as the data collection capabilities of federally funded labs and centers improves, how are campuses helping to manage productive access to this data? Finally, new compliance requirements for security and privacy will affect approaches to managing the data.
  
  
• Connectivity across campus and to external networks for these compute servers: For centralized servers, there are issues associated with provisioning outputs to the visualization engines and other post processing that are usually back in the scientists’ laboratories. For external connectivity, advanced networks and direct lambda connections pose interesting options and security challenges.
  
  
• Building a cooperative effort: Frequently, a campus will seek to leverage one or several computing grants given to individual research groups to create a campus computing coalition. Approaches to building these coalitions, in areas such as proper machine environments (especially for space and environmentally constrained departmental labs), greater computing capacity opportunities, and security need to be identified and verified.
  
  
• Management issues: The ongoing costs of maintaining these environments, particularly for specialized hardware or operating system/grid layer software, need to be allocated, either to users, funding agencies, or campus IT management. Management must also address the periodic need to synchronously upgrade the computing service, regardless of whether the subsystems were acquired piecemeal over several years. And, there is the need to establish allocation of the computing resources themselves to the subscriber community.

The workshop is planned to be adjacent to the Spring Internet2 Member Meeting, April 25-27, 2006. The first half day is targeted to both campus CIOs and CTOs, while the remainder of the sessions are aimed more at CTOs and those more directly charged with campus research computing support.

This workshop is supported with funding from Internet2, Pennsylvania State University,
and the NSF - Grant No. OCI-0627970. Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation (NSF).