VIAS MySQL Perl Engine nuggets for each metadata topic Internet Search fails in Schools Why not use Google? 10.30 --> tomorrow Partners Pittsburgh SDSC NCSA together on DTF Blue ribbon panel Paul Messina Jerry Ostriker 4 sub-panels including applications technology December report Cyber-infrastructure partners to ensure cutting edge CS applications synergy (both ways) CS Infrastructure linkage Contradiction No money to harden software Grid AT (physics) Condor AT Need AT to match technologies What learnt Engagement with applications? MRE Major Research Equipment DTF Application software initiatives like PET Paradigm shifts Grid What is next one? Decoupling partnership and facilities "New young scientists" Software house for Application community Shift to Software Split Partners facilities dangerous EOT combined is succesful McRae -- needs lots of consultants Not computer science community NRAO (VLBI) NO users!!!!! Need users Keep AT teams Evaluate ET people Need AT/ET Interface people What is Organization Bring AT and ET together is mission Pittsburgh couldn't invent a grid! Alliance does real projects well bad at indefinite projects Sorties -- Iway Permanent settlement -- Globus Wny no AT-ET interactions Problem oriented ------------------------------- City of Mind likes Complex Systems ------------------------------ Dan Reed MRE Distributed Instruments and Scientists ALMA (astronomy) EARTHSCOPE ICECUBE (Neutrino) LIGO NEES Support it www.sims.berkeley.edu/how-much-info Surface Web 25-50 terabytes Deep Web 1 Petabyte -- NASA NCAR 13.6 TF Linux Teragrid no Java Metadata management SRB Hotpage Nation Scale multidisciplinary problems PetaGrid Teraflop for $50K Microgrid Power management Quaatum Computing? Gannon ET 20 people Large Scale Clusters X in abox Globus and Condor Crutcher Community Codes ****** Work with ET Test and Deploy new infrastructure Science Community Code in a box **** PACS Gilfeather stepped down Application Portals Support DTF Obviously a problem Harvey Newman -- Condor POrtal Giles Access Grid Clusters EOT Success Nanomaterials is a big national initiative Blue Gene is a thrust not a machine Life Sciences is IBM initiative Need users from MRE EOT and DTF Need democratic grid ------------------------------------------ Discussion on Day 2 My comments: Focus on partnershp. Mission: bring computer science and application communities together to make better science and computer science for HPCC? But what is the right model? CS enables science? AT team strategies automated? Issues: politics, management, new technologies, software, Politics: Will there be an Alliance? Will the PACI program be unified? Will the partnership be separated from the facilities and unified? Management, what is the proper role of the collaborations (ET and AT)? How can we use management and budget strategies to further those collaborations (ensure they work) Technologies: What are the sorties in the future we should come together to enable? How can we identify these and lead their development? How can we establish permanent communities on the beachhead. Software: How can we automate some of the strategies used by AT teams to How can we capture the ideas in robust software, both tools and community codes? Here the issue is leverage? Can open source help? PACS problem Cant commercilaize Can you commercialize open source Need a $50m/year software program Access Grid Process worked as finite problem mailing list good finite time Infrastructure good but Alliance current work good and succesful Defining Endorsing computational science Methodology for world ASCI DoE NASA Access Grid in a box HDF added Grid Forum DoD/PET JPL Earthquake simulation community PSP Princeton -- more and more users Start up packages -- less space more money Not supercomputer users Average user -- can learn too much bureaucracy access Better publicity 1/3 of machine to small users Use Princeton Idle cycles Gregg V Universities are good at somethings but NOT Q/A Must play on strengths Mary V Distinguish from next generation software Reed Infrastructure takes time Who else can address petascale computing McRae Restructure funding Software Engineers Industry link Linux not important there Kennedy million lines of code in his compiler Can't help Woodward McRae Translate Leading edge --> rest of users Number of women and minority in team McRae Software lasts 10 years Hardware lasts 1 year (half lives) Needs seamless mapping PACS Relationship PSP Relationship A) Software Infrastructure Domain PSE: community code and systems Optimization Data Assimilation B) Working with MRE what are futures how do we service community Bring Resources to the users computational science in a box scientific american Need to accelerate petaflops Areas needing tools -- Livny doing GLOW -- Wisconsin grid ---------------------------------------------------------- EOT Collaboration in a box Distance Education in a box Peer to Peer Grids ---------------------------------------------------------- No animation in Access Grid ------------------------------------------------------------ NEES is a Systems Integration Project Soil Foundation <--> Structures Interactions wireless MEMS Instrumentation lots of data GryPhon Sloan digital sky survey LIGO Atlas CMS Tier0 --> Tier4 CERN ---> Desktop Metadata catalog Condor Scheduling AIPS++ Radio Astronomy VLA Greenbank Altacama .... Altacama large millimeter radio $700M NVO National Virtual Observatory AIPS++ 150 FTE (parallelized by NCSA) COBRA Radio Astronomy Data Archive --------------------------------------------------------- Canarie -- fiber is a storage device Grossman 10 gigabytes per person 1999 > Exabyte storage sold Deep data -- data web 16 terabytes of conventional web now petabytes of data web Stevens working with BUAA IBM Toshiba Sony Play Station 3 thousand times PS2 Design center in austin Linux is OS for PS2 Biology community is huge CS is too small Digital database of human knowledge wants broad grids (education) VR SI Roadmap OK till 2010 -- what about after this "human creativity" Software is limitation Physicists are reductionists Engineers Integrate (study systems) Need women speaker Wars ---------------------------------------------------------- Kennedy May 25 2001 Gregg: Generic Services -- Abstract user interface User Survey Petaflop leader or DoE follower ASCI conservative -- not interested in Grids Now ASCI gets lots of software from NCSA Reputation rests on success of Grid Tiger team at Alliance to help application people write Cactus is application independent toolkit AIPS++ is similar Application middleware Take Portals to logical conclusion Currently math libraries etc. provided by IBM SGI Needed to be done for Grid Jerry Ostriker -- get extra people from hardware! NLANR Distributed Application Support Team not so succesful ---------------------------------------------------------- NSF Hildebrandt PSC is a managemenr problem 3 site team www.paci.org Hot Page OpenMP MPI online material Langer report (downloaded) Altacama Large Millimeter Array ALMA NEON National Ecological Observatory Network CyberInfrastructure integrates Push in each discipline Blue Ribbon Panel ==== Advisory Committee on Cyberinfrastructure b_ribbon.html under NSF web site http://www.cise.nsf.gov/ciseac/acreport_2001_04.html ------------------------------------------------------------- Participatory design of technology Roscoe Giles