SUG Meeting Minutes January 12, 2006

Welcome / Robert Marcus

• Members moved and approved agenda
• Members moved and approved minutes from last meeting
• SUG Executive Committee Meeting
  • Broaden base of SUG
  • Guest speaker series possibilities (national speakers or annual distinguished lecture series. Asked SUG members to give speaker ideas before they leave meeting today
• Ralph Regula School of Computational Science: Asking SUG members to help get the word out

Software Update / Rick Prairie

• Portland compilers license upgraded to include Opteron support
• Altair Hyperworks renewed
• John Hackett of OSU asked upgrade Turbomole from 5.71 to 5.8. Received quote for upgrade and maintenance for OSU Chemistry, OSU Campus, and all Ohio. Discussion that there are users on campuses other than OSU
• MATLAB may require non-OSU users to use DCE to run at OSC using licenses defined on their home workstations. Sam thinks current license contains all toolkits – all inclusive
• Leslie said we need metrics to help support our software purchases
• Bioinformatics software
• Mascot upgrade 2 CPU + memory Linux conversion. Should be installed tomorrow
• TimeLogic FPGA Upgrades: New board in full production on P4
• Temp reintegrated with Sun portal based on user request
• Still interested in Production use of XD1 nodes for some algorithms
• Providing sequence searching support for selected collaborator websites
• Sequence database downloading. Resuming and working through details of current environment
• Partek Genomic Suite Module purchase. Did evaluation. 4 larger labs definitely use. More labs likely to follow

Allocations Update / Christopher Hadad

• Start up grants: 9 for total of 45,000 RUs
• Classroom grants: 0
• Standard grants: 3 full for total 30,000 2 partial awards 2 for total 10,000
• Major grants: 3 full and 2 half for total 120,000 RUs
• Total RUs awarded: 215,000
• Jim did study to compare awarded, active, and available RUs
• From last SUG meeting, allocations vs available RUs Total Positive RUs 1,490,410 Projects 1282
• Total active RUs 1,111,526 (current and active projects) Projects 252
• Available system RUs 1,012,656 (no downtime)
• SUG Exec suggested creation of potentially large awards
• Current limits 5K, 10K, and 30K RUs
• Suggestion of perhaps 250K from one proposal
• Allocations would discuss review criteria and mechanism
• What does SUG think?
• Dick Piccard said that traditionally OSC awarded RUs with no restrictions on how quickly used up. If someone uses too many RUs in a month, can create hardship for others.
• Christopher said ceilings need to be set
• Rick said this is not good thing to do now until we get new hardware
• Anthony Johnson suggested we award percentage of RUs available, not set number
• Someone suggested that users may be going elsewhere due to features of a machine, not RUs
• What's the next step? Weblog approach for feedback

OSC Update / Stan Ahalt

• Compete Columbus group hosted meeting today. Deborah Wince-Smith, Council on Competitiveness, was featured speaker. She mentioned OSC in her talk three times.
• Thanks to SUG for contributions to OSC success
• e-Tech MOU signed. Statewide event at end of February
• OIT negotiations continue
• Confidential NSF proposal discussion
• Introduction of new Chief Operating Officer George Haller. We are grateful that George has joined us

Honoring Austin Melton

• Stan Ahalt thanked Austin Melton for his contributions to SUG as former Chair. He presented him with a plaque
• Current Chair Robert Marcus thanked Austin for his service and legacy and said that he would continue with the initiative to grow the SUG membership

Hardware Update / John Heimaster

• Storage challenge – large growth in home directories. Have to be sure in position to deal with storage growth
• Next major acquisition
• Saturation of clusters with undesirable delays
• Working on RFP (best way to acquire and maintain systems)
• Need input from current users and others
• Arguing for balance across wide range of applications

SUG Member Instructional / Research Updates

• Robert Marcus teaches undergraduates how to program “Intro to Parallel Computing with Animated Algorithms” using Web based material
• He would like to collaborate with other computer science instructors

HPC Update & User Services Update / Leslie Southern

• Leslie just returned from GGF / EPIC Meeting
• EPIC is follow-on to PACI project
• Virtual education and training collaborations
• Six functioning grids in world – teragrid is one of them
• SC 05 Highlights
• Active participation among OSC industry and academic partners in technical and exhibit programs
• Booth presentations focused on collaborations and demos
• BOFs, Tutorials, and Workshops
• Council on Competitiveness, chaired by Stan Ahalt, one of best attended BOFs at SC
• Interactive Computing with MATLAB
• Open FPGA
• GridChem
• Response to inquiries: Governing Board requested external funds to be attributed to statewide faculty using OSC
• 22 institutions contacted. 11 responded
• 76 sponsors, 240 grants, $21,541,426
• Based on data received to date and research funds by state universities, we estimate $60,000,000 for OSC leveraged project funds to statewide faculty
• Hardware Installations / GRI/HLL Research Cluster at Springfield
• 84 available nodes (Dual AMD Opteron, 2.2 Ghz)
• #2 8 GB (front-ends / interactive use)
• #72 4 GB nodes
• #8 8 GB nodes
• #2 16 GB nodes
• Software stack of bioinformatics applications
• Research for parallel MATLAB development
• OH Board of Regents, Genome Research, and Higher Level Language Research Collaboration
• HPC Acquisition Update
• Looked at number of different architectures and applications
• Broad user community impact with the best price / performance
• Different sizes of nodes based on workload
• Vendor configurations based on three levels of funds
• Example configuration that would be in the Top 50 of the Top 500 list
• 768 node, 3328 (x86_64) processor cluster
• 704 dual chip/dual core systems with 8GB memory
• 64 quad chip/dual core systems with 32GB memory Infiniband 4x DDR interconnect ~3us latency ~1600MB/s bandwidth 14.64TFLOPS peak FP performance
• Consultation Metrics
• Using database to track consultation logs. Can track how long tickets open, what is done, types of problems etc.
• General Unix / System, systems staff, programming, third party software packages, batch, and other
• Type of consultations changing. Some users now don't know UNIX because they are working in a GUI world
• Training
• December
• 13 & 14, Parallel MATLAB; at ASC
• January
• 11, HPC Overview; at OSC
• 18 & 19, Computational Chemistry; at OSC: Scott Brozell providing instruction
• February and March
• 1 & 2, Python Workshop; at OSC
• 15 & 16, Introductory and Intermediate MATLAB; at OSC
• 2/28, 3/1 & 3/2, Software Development Tools; at OSC: new workshop tools people need to manage these software projects
• Potential to stream future workshops
• Summer Institutes
• Expanding YWSI: Have program at OSC and Kent State University. Will add one at Wittenberg University
• Seeking Funding Opportunities for both SI and YWSI (Corporate Sponsors, Foundations, Government)
• Summer School on Computational Materials Science hosted by Materials Computation Center at UIUC. Approached OSC for instruction
• Provide some instruction for Central State SI

OSC Networking Update / Pankaj Shah

• Showed map representing footprint of current fiber
• All Higher Education institutions are connected to the backbone
• 17 of 17 Phase 2 institutions are directly connected with fiber-based connections
• OSC Springfield circuit being tested
• Youngstown – July 2005
• University of Dayton – August 2005
• Central State has ordered fiber-based circuit from SBC – August 2005
• Kent State and NEOUCOM – November 2005
• 11 Phase 3 institutions are directly connected
• Last Mile Status / Out of 33 institutions with high bandwidth last-mile connections:
• First 17 funded by state funds
• 2 funded by a NSF High Performance Network Connection grant
• 5 to be funded by federal funds obtained by the Ohio Board of Regents
• 9 self-funded connections
• 54 unfunded
• 12 have DS3 and above
• Others have T1s
• TFN Robustness
• October 13, 2005: A pre-amplifier card malfunctioned at the Springfield PoP (Point of Presence). Due to redundancy, no TFN customers were affected while we dispatched an Engineer with a replacement part
• October 19, 2005: There was a massive power outage at our PoP in downtown Columbus. While some other providers in the area lost service, our redundant power supplies and architecture avoided service interruptions
• October 25, 2005: There was a fire that destroyed cable box outside of Canton TFN PoP. As we had fiber redundancy, no TFN clients were affected. It took close to 10 hours for the fire damage to be repaired
• Enhanced ability to monitor TFN allows proactive resolution of network issues
• December 22, 2005: The local loop to NEOUCOM went down due to a transponder failure. The problem was diagnosed in under 15 minutes and the problem was resolved within 90 minutes; diagnose and repair times exceeded industry standards
• December 30, 2005: An AEP transformer exploded and melted fiber and Ring 3 operated as expected, no TFN clients were affected
• Events
• CORN event (Central Ohio Research Network, City of Dublin) August 31, 2005
• Keystone conference, October 4 & 5, 2005
• TFN event at Shawnee State, October 13, 2005
• Megaconference, December 2005
• ODCE / Windows on the Future, March 2006: SUG members invited. Will receive e-mails about it.
• Outreach
• IHETS, September 2005
• MERIT, September 2005
• Agree to work with Video projects
• Work toward Disaster Recovery
• Look toward RON-to-RON peering
• Internet2 Meeting, September 17-22, 2005
• “Broadband 101” at Wooster – ODOD and Whiteboard, Inc., October 28, 2005
• StateNets, August 09-12, 2005
• RONS (Regional Optical Networks)
• SC 2005, November 11-19, 2005
• Internet2 and NLR – Resolution of merger and group reports at http://www.internet2.edu
• CALEA (Communications Assistance for Law Enforcement Act) November 09, 2005
• It is not clear how it will effect Higher Education. The actual requirement has not been defined.
• It is important that Higher Education get involved early and monitor the situation.
• Higher Education is seeking relief from the FCC. OSC Networking has filed a request to the FCC to exempt Higher Education
• It was stated in the request that OSC Networking and Higher Education recognize the legitimate need for surveillance
• OSC Networking has always cooperated with authorities within the requirements of the law in a timely fashion
• Collaborations Update
• Cincinnati Education Research Fiberloop (CERF)
• Dublink (CORN – Central Ohio Research Network)
• OneCleveland
• e-Tech
• OIT
• Showed map of Higher Education Regional Optical Networks Plus Internet2 Plus NLR
• ITIL
• Four modules of immediate interest
• Service desk
• Change management
• Asset management
• Configuration management
• Vendor visits

Closing Comments / Robert Marcus

• Cake honoring Austin Melton in BALE Lobby
• Please give Leslie Southern names of suggested speakers for future SUG meetings