DoE Networking Thomas Ndousse-Fetter head of networking at DoE Office of Science carlson@er.doe.gov Mathematical, Information and Computation Sciences includes collaboratory tools $116 M Base Program $50M New Start Scientific Discovery Through Advanced Computing "High Speed Networking Corridor in SE" Need Metacomputing to solve DoE problems Enhance existing network Need to realize high performance Multi gigabit rates Tera scale data intensive applications / remote visualization Three Components Basic Research in areas like QoS, TCP OS by pass, cluster computer interconnects Middleware Development and Integration Grid enabled Services e.g. network enabled collaboration tools, multicast, security, application aware network services (electronic notebook, globus mentioned under collaboration) Advanced Deployment and Testing killer applications, test new technologies such as IPV6, performance integrate into scientific apoplications Need 3 to 4 facilities linked together e.g. test deployment of PKI UT Battelle is a LLC involving sundry universities and Battelle NSF Interested ORNL wants to be leading High Performance laboratory Open unclassified research Zacharia Computational Material Scientist SDAC: Scientific Discovery Through Advanced Computing $50M CS $20M Applications Follows on SSI -- SSI mainly hardware New program is more balanced ESNET3 ORNL Capabilities material science climate simulation (Malone LANL involved) Computer Science: Electronic Notebook, NetSolve, M3C Cluster Management Tool Largest unclassified computing facility IBM 1 Teraflop, Compay .5 Teraflop IBM ahead in software Uses TOP 500 plots to plan performance needs Batelle runs/part of management at: PNNL ORNL NREL NETL UTK-ORNL: Joint Institute for Computational Sciences Expand to other universities Joint research or appointments Software Challenges Currently legacy software like DYNA Lets restart from scratch Systems software Scientific Algorithms New architectures Teams linking CS and application types Porting Codes Petabytes of Data Investments designed to maximize scientific results Napear (Outreach lead Academia--Industry from 1990) napear@sdsc.edu Storage capacity increasing as fast as communication bandwidth Today: Machines to people dominate Tomorrow: Machine to Machine (Computer in Sensor) in 3 ways Environmental Data is "cold war" of next century in that will occupy energy of the world Ground Sensors e.g. smart houses, sewers, bridges etc. Health Sensors Link Sensors to Simulations DAta mining etc. SURA Gary Crane Doe connection from Jefferson Lab Money ($3M IT Initiative fund) 3 Current Projects Regional Network Infrastructure Video Projects VIDE has a cookbook of how to use video conferencing Gigapop community ISP services SURA IT Model versus DoE Science Grid SURA model seems very organizational as opposed to technology based Very confused discussion of Grid and user interfaces Should you put grids in a university? SURA and ORNL Relationship SURA funds workshops Seed funding for projects ORNL (re)join SURA? Qwest US Europe Technology Summary Multi-technology bridges are expensive pure optical networks only require regeneration every 1600 km DARPA BOSSnet MIT -- Washington DARPA Supernet across country Abilene: 130 institutions linked in Internet2 Visual History Foundation Doe LBNL LLNL ANL ORNL Super-Regional Network OC 48c Academic / Research Institutions Indiana--Washington--FSU Link to Chicago, ESNET Ron Hutchins Georgia Tech Infrastructure: Regions Good Hierarchial Organization Special purpose regional resource Earthquake Sciences NEES Solicitation Dan Stevenson MCNC Applications Multi-video Office HDTV Virtual Surgery Real time meteorology Real Time Air quality -- Traffic simulations Very impressive network Network research High apeed network security optical network Possible Research Areas Optical Protocol Advanced IP Mobile Protocol ATM and MPLS Intrusion detection/security Network Simulation Remote Visualization Data Handling Dynamic routing and bandwidth brokering ESNET3 at Oak Ridge Network for Basic Research different from Applications Crash network or not? As soon as network works users demand it become production Jefferson Lab is a source of data Also Oak Ridge Morphnet switches between modes back up network is research Users want to use network as an appliance Partitioned infrastructure Need applications but not production .... Fault tolerance research helps to hide problems needs flaky computers and flaky networks Central Activity -- Network plus "ORNL Organization" Name: Network Systems Research Center for Deep Networking Center for Communities of the Future ------------------- May 17 ---------------------------- SNS will give a lot of data -- Spallation Neutron Source Network System as an Instrument Use Network Instrument Develop Network Instrument Distributed Scientific Computing Virtual Research and Education Communities Aggregation and Access to Community Resources Experimental Network -- Network as an Instrument Marshalling of Network Resources Applications are in SDAC Telemedicine is NIH not DoE Network Instrument can be used in NIH proposal ... Motivation for Regional Network Network as an Instrunent Value of Integrated approach Research for the Instrument Advanced IP etc. Research on the Instrument Distributed Scientific Computing, Virtual Research and Education Communities Why this is the right time Capabilities and strengths of SE region Resources on the Instrument Research Expertise Potential Impact and Enabling Technologies Industrial Support Management Experimental Network is an Instrument which supports research on network devices as appliances Total is 10-20 pages Deepnet is instrument South East Networked Systems Concortium Global (Application) Challenges What can you with distributed teams in education and training Building communities with robust interactive environment What are requirements for various communities network device as an appliance metacomputing (low latency high babdwidth) network sensitive applications Hand held to Cave -- universal access Technical Challenges Portal / Collaboration Architecture Virtual Environments Integration of network monitors into applications Grid Technology (global MPI, Security) New models like Jini Audio Video Conferencing Collaborative or Remote Visualization Seamless Access