Computing - Department of Physics - University of Illinois at Urbana ...

PHYSICS ILLINOIS NEWS • 2006 NUMBER 1 3Larry Smarr: NCSA’s Architect Still Revolutionizing Information InfrastructureLarry Smarr as a young assistant professor,when NCSA was only a dream.BY CYNDI PACELEYWhen he left the University ofIllinois for the University ofCalifornia six years ago, former Physicsprofessor Larry Smarr knew thatuncharted waters awaited him. But littledid he realize that he would also gainthe opportunity to captain a maidenship and to learn about entirely newacademic disciplines—including a keycomponent of our earth’s oceans.New frontiers have always appealed toSmarr, who knew from the first gradethat he was going to be a scientist. Earlyin his career, Smarr made his mark inrelativistic astrophysics, founding thefield of numerical general relativityand making major contributions tocomputational high energyastrophysics—supernovae, neutronstars, blackholes, relativisticmagnetohydrodynamics, and galacticjet dynamics.It was his work in computationalscience that led him to realize the needfor an integrated system for computing,storage, visualization, and analysis tosupport scientific research.Not surprisingly, Physics played amajor role in the origins of such asystem. With a joint appointment inPhysics and Astronomy, Smarr and hiscolleagues began building an earlyversion of a supercomputing center in1979–80. With the fitting acronym ofVIP (VAX and Image ProcessingCenter), a consortium of faculty inphysics, astronomy, atmosphericsciences, and theoretical and appliedmechanics pursued interdisciplinarycollaborations in computer simulationsand visualizations of large data sets.“It was a first step in gainingexperience to address the very limitedcomputing power available to universityresearchers at that time,” Smarr recalled.Creating VIP sparked discussionsacross the Urbana-Champaign campusabout supercomputing’s increasinglycentral role in basic research. At thesame time, Smarr was exposed tosupercomputing in action. As a Fellowat Germany’s renowned Max PlanckInstitute for Physics and Astrophysicsduring the summer of 1982, he sawfirsthand what a Cray supercomputercould do—and that its speed was severalhundred times faster than anythingachieved with the local VAX computer,which was as large as any US researchgroup had at the time. These experiencesled Smarr and his UI colleagues tocompile 65 case studies from 16departments, demonstrating the needfor greater computing capabilities tocarry out research projects.Putting It In WritingIn 1983, while the National ScienceFoundation was still conducting blueribbonpanel discussions on the need forsupercomputing, Smarr wrote a proposalrecommending development of anational supercomputer center.“It still amazes me that I—an assistantprofessor at the time—had the nerveto author an unsolicited proposal for$55 million to the NSF,” Smarr said.He sought the guidance of RalphSimmons, head of the physicsdepartment, on how to best deliverthe document.“Ralph suggested that we put it in anenvelope marked ‘Director, NSF,’ hop ona plane, and deliver it to NSF’s Math andPhysical Sciences assistant director MarcelBardon, whom Ralph knew well,” Smarrexplained. “That someone of Ralph’sstature was willing to support a juniorprofessor like me in such a substantialway speaks volumes about hisleadership.”After a “frustrating” wait of a year anda half while the NSF leaders reviewedSmarr’s proposal as part of a new nationalcompetition, the University of Illinoiswas tapped as one of the nation’s fiveNSF supercomputer center locations,creating the National Center forSupercomputing Applications (NCSA)in 1985.But the first step was finding a suitablelocation for the actual supercomputingunits. Enter another Smarr mentor inthe person of Ned Goldwasser, a formerPhysics professor who had becomeFermilab’s deputy director in 1967and then returned to campus as vicechancellor for research and dean ofthe Graduate College.“Because of his background in physicsaccelerators, Ned understood that weneeded a world-class facility to house thesupercomputers and their storage units.Fortunately, the Advanced ComputationLab’s second floor had been built tohouse the University’s ILLIAC IVcomputer in 1968,” Smarr said. “I toldNed that with a little work, it couldbecome our first supercomputer workroom.” That “little work” translatedinto several million dollars—a hugecommitment for the campus at that time.It was quickly apparent that the firsttwo-processor Cray X-MP would soonhave to be upgraded, since a lonetheoretical and particle physicist—JohnKogut—could easily use all of one of theprocessors while the other was reservedfor the rest of the country.“I then worked closely with TedBrown, the vice chancellor for research,and George Badger, head of theComputing Services Office, in creatingthe structure of NCSA. By that time,I was so far into it, I had to become thefirst director,” Smarr said. He assumedthe reins in 1985 and was namedin 1997 to head the NationalComputational Science Alliance beforestepping down in March 2000.“A Long, Proud Tradition”Smarr saw NCSA as “a very naturalstep in a long, proud tradition ofsupercomputing technology for theU of I at Urbana-Champaign.”In answer to those who attempt tocredit solely him for NCSA, he offersan interesting historical footnote tocoincidence or fate—or both.“A letter dated October 16, 1948—the day I was born—was sent from theU of I to J. Robert Oppenheimer, whohad recently left the Los AlamosManhattan Project to direct the Institutefor Advanced Study in Princeton, NewJersey. The letter advocated plans forconstructing two “supercomputers”—one in Princeton for John vonNeumann and the other at the U of I.That computer became ILLIAC I.“So I can say that fully all of my life,the U of I has been at the forefront ofhigh performance computing,” Smarrsaid.“Completed 31 years before my initialNSF proposal, ILLIAC I was that keyfirst step in making Illinois a computingpowerhouse,” he continued. “As a resultof the ILLIAC I, II, III, IV, andCEDAR, the University for manyyears recruited faculty in differentdepartments who did the type of worksuited to supercomputing,” he added.“That Illinois had the fertile intellectualground already in place—and nationallyrecognized—was the reason we wereable to organize around NCSA.”Heading WestPursuit of emerging science has been oneof the happy outcomes of Smarr’s moveto California in 2000, where he wasrecruited to the University of California,San Diego (UCSD) in the Departmentof Computer Science and Engineering.Although intending to return to basicresearch in distributed computingarchitectures, Smarr was again asked toplay a trailblazer role—this time toanswer the call to create a CaliforniaInstitute for Science and Innovation,an initiative championed by both thestate’s governor and then UC PresidentRichard Atkinson. Smarr was asked topull together a team between UCSDand UC, Irvine, develop a proposal,and win the competition.Given his previous success in launchingNCSA, it’s no surprise that Smarr did justthat. Bringing together more than 100faculty at the frontiers of informationtechnology, telecommunications,nanotechnology, and applications of all ofthese to biomedical research, intelligenttransportation, the environment andcivil infrastructure, and the new mediaarts “was a once-in-a-lifetimeopportunity,” he said.The Institute’s newest project, justunveiled in mid-January, unites Smarr’sprofessional interest in computationalstructures with a personal love of marinereefs.Researchers at UCSD will build a stateof-the-artcomputational resource anddevelop software tools to decipher thegenetic code of communities of microbiallife in the world’s oceans. The newresource will help scientists understandhow microbes function in their naturalecosystems and enable studies on theeffect humans are having on theenvironment, as well as permit insightinto the evolution of life on Earth.The project gives Smarr “a wonderfulopportunity” to learn entirely newdisciplines, such as microbial ecologymetagenomics “with mentors who arethe worlds’ leading scientific researchers.”“I get to combine that frontierscientific-discovery excitement with theability to architect whole new classes ofscience servers connected by dedicatedoptical networks to the laboratories ofend-users around the globe,” he said.“It’s hard to imagine a more excitingchallenge.” ■Announcing the Moore Foundation grant for the CAMERA project. Left to right: CraigVenter, President, J. Craig Venter Institute; UCSD Chancellor Marye Anne Fox, Calit2Director Larry Smarr; and Scripps Institution of Oceanography Deputy Director John Orcutt.1951 Hoping to avoid the mass exodus of Urbanaresearchers that occurred during World War II, Louis Ridenour,now dean of the Graduate College, works with Wheeler Loomisto establish the Control Systems Laboratory (CSL). Funded bygrants from all three branches of the Armed Services, the CSLfocuses on radar and on possible uses of the new digitalcomputers.1957 The Digital Computer Laboratory,which has been offering classes tostudents since 1954, becomes a fulldepartment with Ralph Meagher (Physics)as its head. (A subsequent head, James N.Snyder, will also come from Physics.)1959 Daniel Alpert(Physics) begins aproject in CSL forcomputer-assistedinstruction, which willcome to be known asPLATO.1960 A patternrecognition computer(ILLIAC III) is designed inPhysics to analyze bubblechamber photographs ofhigh-energy particleevents.1961 Using ILLIAC I as a computational engine,PLATO becomes the nation’s first computer-assistedprogram of instruction. Conceived by Chalmers Sherwin(Physics) and developed under the direction of DonBitzer (Electrical Engineering), PLATO is the world’s firsttime-shared computer-based education system and thehome of the world’s first on-line community.UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGN