Docstoc

DOE net projects - No Slide Title

Document Sample
DOE net projects - No Slide Title Powered By Docstoc
					Department of Energy                                        Office of Science
                                                              MICS Division



                         Project Quad Charts



              High-Performance Networking Research Program

                        Program Manger: Thomas D. Ndousse
                                 Tel: 301-903-9960
                            Email: tndousse@er.doe.gov
                  Bandwidth Estimation: Methodologies and Applications
                               k claffy, CAIDA at SDSC & Constantinos Dovrolis, Univ. of Delaware


   High-Performance Network Research                                                   SciDAC Project
                                                                                 The Novel Ideas
           Brief Summary of the Project                          Innovative end-to-end probing techniques to measure
 • Task 1: Develop accurate, fast, and non-intrusive            capacity (max possible throughput in empty path) and
 bandwidth estimation (bwest) methodologies and                 available bandwidth (max throughput under current load):
 measurement tools.                                                    – Packet Train Dispersion (PTD).
 • Task 2: Compare and evaluate different bwest tools                  – Variable Packet Size (VPS).
 (both for end-to-end and per-hop bandwidth metrics),                  – Self-Loading Periodic Streams (SLoPS)
 characterizing any observed errors.                             Methodologies to check for overbuffered or underbuffered
 • Task 3: Use bwest methodologies in transport                 network paths.
 protocols and applications to optimize throughput for           Smooth pacing in TCP, driven by bwest measurements.
 high bandwidth-delay-product paths.                             Smooth bwest driven rate-control for UDP-based
 • Task 4: Prototype bwest middleware to monitor                applications.
 performance between network domains in real-time.

           Impact and Connections                                              Milestones/Dates/Status
 IMPACT: Allow scientific applications (transferring            Compare and evaluate existing bwest tools:
terabytes of data) to efficiently use high-performance              - Hop-by-hop tool survey                   Jun01 - Aug02
networks .                                                          - End-to-end tool survey                   Jun 01 - Jun02
        – Use explicit bwest measurements instead of             Bandwidth measurement middleware
        implicit bwest via TCP’s congestion control                 - Create/maintain testbed                   Jun01 - Jun04
        algorithms.                                                 - Collect link characteristics             Jun01 - Jun04
        – Provide easy-to-use tools for monitoring network          - Correlate active/passive measurements    Jun01 - Jun04
        path performance.                                        Capacity estimation tool (pathrate) v2.1.2    Dec01 DONE
CONNECTIONS:                                                       - Add GUI to aid analysis of results                Dec02
        – Apply bwest methodologies to Web100 and                Available bandwidth tool (pathload)                  Mar02
        Net100 projects.                                            - Paper at PAM’02                                  Mar02
        – Correlate bwest to loss/delay (e.g. PingER project)    Develop UDP-based rate-controlled file transfer app
        – Establish prototype bwest middleware in ESnet         driven from bwest measurements                          Dec02
        and for DOE labs and investigators.                      Real-time path monitor using bwest middleware          Dec03

                             MICS Program Manager: Thomas Ndousse                                              6/19/2010 8:59:41 AM
                    Security and Policy for Group Collaboration
                Steven Tuecke, Argonne National Laboratory, Carl Kesselman, USC Information Sciences Institute
                                                      Miron Livny, U. Wisconsin, Madison
      High-Performance Network Research                                                                                                    SciDAC Project
                                                                                                            The Novel Ideas
                     1. CAS request, with                               user/group
                                                    CAS
                                                                                          Enable collaborative work, with common security tools that address:
                        resource names                                  membership
                        and operations             Does the
                                               collective policy
                     2. CAS reply, with
                          capability
                                                authorize this
                                               request for this
                                                                   resource/collective
                                                                      membership            - Large, geographically & organizationally distributed membership
                        and resource CA info         user?
                                                                     collective policy      - Membership with diverse expertise, comprising different roles
                                                                       information
                                                                                            - Community resources with associated community policies
             User
                     3. Resource request,
                        authenticated with
                                                Resource                                  Develop novel tools and approaches for:
                          capability           Is this request
                                               authorized by
                                                                                            - Management of collaboration membership and resources
                                                                                                 - Online CA & Credential Repository (CR), local security integration
                                                      the
                                                 capability?            local policy
                                                                        information
                                                                                            - Management of roles and privileges
                     4. Resource reply

                                               Is this request
                                               authorized for
                                                 the CAS?                                        - Community Authorization Service (CAS), restricted delegation
                                                                                            - Integration into collaborative tools and environments
                    Community Authorization Service

           Impact and Connections                                                                  Milestones/Dates/Status
 IMPACT: We expect this project to result in:                                            Demonstrate CAS prototype @ SC’01                    November 2001
      Standardization of new PKI-based approaches to credential                          Complete X.509 & GSS standards drafts                February 2002
       management, restricted delegation, policy management                               Deliver draft standard conforming GSS                April 2002
      Development of security tools and services for collaboration                       Deliver CAS w/ simple policies                       May 2002
      Widespread deployment and adoption of approaches and tools
                                                                                          Demonstrate Online CA & CR                           September 2002
 CONNECTIONS:
                                                                                          Complete Online CA & CR standards drafts             December 2002
      This work builds on the Globus Toolkit’s widely used Grid
       Security Infrastructure (GSI), and will be in future Globus Toolkit.               Finalize X.509 & GSS standards                       February 2003
      To be used by numerous SciDAC collaboratories, including DOE                       Deliver Online CA & CR                               March 2003
       Science Grid, Particle Physics Data Grid, Earth Systems Grid,                      Deliver CAS w/ rich policy & app support             May 2003
       and Fusion Collaboratory                                                           Finalize Online CA & CR standards                    December 2003
      Also to be used by many non-DOE projects worldwide, including                      Deliver standards-based Online CA & CR               March 2004
       NSF PACI DTF, NASA IPG, and European Data Grid                                     Deliver CAS w/ accounting support                    May 2004


                                                                                                                                                    September 2001
                    MICS Program Manager: Thomas Ndousse
                  INCITE: Edge-based Traffic Processing and Inference for High-Performance Networks
                               Richard Baraniuk, Rice University; Les Cottrell, SLAC; Wu-chun Feng, LANL

           High-Performance Network Research                                                                      SciDAC Project

                  INCITE Summary                                                     INCITE Novel Ideas

• Task 1: Multiscale traffic analysis and modeling                • Multiscale / multifractal analysis for traffic bursts
• Task 2: Inference algorithms for network paths and links        • Efficient “packet chirp” and “fat boy” path probing
• Task 3: Network tomography                                      • Active and passive network tomography
• Task 4: Active network measurement: PingER                      • Monitor for Application-Generated Network Traffic (MAGNeT)
• Task 5: Passive network Measurement: MAGNeT, TICKET
                                                                  • Traffic Information Collecting Kernel with Exact Timing
• Task 6: Passive path monitoring and tomography toolkit
                                                                    (TICKET)
                    incite.rice.edu                               • Augmented PingER


            Impact and Connections                                                         Milestones
   IMPACT:                                                         Analysis, modeling, and inference
       Optimize performance of demanding applications                      Multifractal, wavelet, tomography theory           ongoing
        such as remote visualization and high-capacity                      Traffic analysis toolbox            12/02
                                                                            Passive path inference and tomography algs           10/03
        data transfers
       New understanding of the complex dynamics of               PingER
        large-scale, high-speed networks                                   Add tomography, chirping, fat boy                     04/02
       New edge-based tools to characterize and map                       Port extended PingER to Rice/LANL                     10/02
                                                                           Add new inference algs to PingER-NG                   06/03
        network performance as a function of space,
                                                                           Evaluate, port PingER-NG to GIMI/NMF                  04/04
        time, application, protocol, and service
                                                                    MAGNeT / TICKET
   CONNECTIONS:                                                          MAGNeT, TICKET (alpha distribution)                     10/02
       Rice/SLAC/LANL synergy, SciDAC                                    High-speed, high-utilization traffic traces             09/02
                                                                          MAGNeT (public availability)                            06/03

                                  MICS Program Manager: Thomas Ndousse                                               Date Prepared: 10 Jan 02
                                            Logistical Networking
              PIs: Micah Beck, Jack Dongarra, James S. Plank / Tennessee; Rich Wolski / UCSB

High-Performance Network Research                                                                              SciDAC Project
                                                                                         Novel Ideas
      Logistical Networking: Developing a
        communicative infrastructure with persistence                Storage is too cheap to hoard.
                                                                     Storage can be a scalably shared network resource.
  Tasks:                                                             Logistical Networking gives applications and middleware uniform
                                                                    control over buffering and routing of data.
  -develop/deploy network storage depots
                                                                     Data storage and data transport can be viewed as points on a
  -develop layered storage stack & tools                            spectrum of data management mechanisms.
  -develop/validate scheduling techniques                            Monitoring and prediction can replace reservation as a means of
  -optimize application performance                                 scheduling storage resources.
                                                                     End-to-end networking principles can apply to storage.
                      loci.cs.utk.edu

         Impact and Connections                                              Milestones/Dates/Status
 IMPACT:                                                         6-12mos –IBP applications demonstrated at SC’01
      Improved performance and scalability of data-intensive              –exNode support in NetSolve
       distributed application                                             –Reliability/performance coscheduling alpha
      Greater ease of and lower cost of deployment of new wide            –Allocation policy simulation
       area data management strategies
                                                                           – Initial generalized caching infrastructure
      Dramatically improved flexibility in data-intensive        12mos
       collaboration                                                       –Initial logistical overlay network on ESNet
 CONNECTIONS:                                                             –Wide-area logistical peering mechanisms and policies
                                                                  12-18mos
      SciDAC: Net100, Data Grid, Scalable Systems, Data Mgt,              –Resolution for highly volatile storage resources
       Computational Science (e.g. Climate, Supernovas)           18-36mos –Experimental IBP architectures
      Base:Network Monitoring, Data Grid, Transport Protocols,            –Large scale measurement and simulations
       Storage Res. Mgt., IQ-Echo,



                               MICS Program Manager: Thomas Ndousse                                                      Date Prepared: 1/10/02
                                                         Net100
                      PIs: Wendy Huntoon/PSC, Tom Dunigan/ORNL, Brian Tierney/LBNL

     High-Performance Network Research -                                                                                Base Project
                                                                                    Net100 Novel Ideas
     NET100: Developing network-aware operating
                            systems                                Net100 will tune network-UNaware applications based on recent
  Tasks:                                                          and current link characteristics
                                                                   Net100 will tune more than just transport buffer sizes, such as
                                                                          TCP AIMD parameters
   -develop/deploy network probes/sensors
                                                                           DUP threshold
   -develop network metrics data base                                      Delayed ACK
   -develop transport protocol optimizations                       Net100 will determine optimal paths and whether to use multiple
   -develop network-tuning daemon                                 streams and/or multiple paths
                                                                   Net100 kernel utilizes passive monitoring from the Web100 kernel
                      www.net100.org

         Impact and Connections                                               Milestones/Dates/Status
 IMPACT:                                                              Network probes and sensors           Mon/Yr     DONE
      increase throughput of bulk transfers over high delay,         - initial sensor and tool deployment    12/01      12/01
       bandwidth networks (like DOE’s ESnet)                          - data base design                       4/02
                                                                      - initial data base implementation       9/02
      select optimal paths and transport parameters for              - final sensor/data base                 6/03
       distributed (Grid) application (e.g.: GridFTP)
      provide network performance data base from active and      •  Transport protocol optimizations
                                                                    - protocol analysis                        11/02
       passive monitoring                                           - initial tuning daemon                     3/02
 CONNECTIONS:                                                      - bulk transfer tuning demos               8/02
      SciDAC: Astrophysics, Bandwidth Estimation, Data Grid,       - final tuning daemon                      6/03
       INCITE, Logistical Networking                               Multipath support
      Base:Network Monitoring, Data Grid, Transport Protocols      - analytical analysis                        8/02
                                                                    - proof-of-principal routing daemons        12/02
                                                                    - grid applications demos                    4/03



                               MICS Program Manager: Thomas Ndousse                                                              Date Prepared: 1/7/02
                             Self-Configuring Network Monitor (SCNM)
                                     PIs: Brian Tierney/LBNL and Deb Agarwal/LBNL

  High-Performance Network Research                                                                                          Base Project
                                                                                                Novel Ideas
     SCNM: Developing a distributed passive network
                        monitoring system                             • A secure monitoring infrastructure that applications can use to monitor
                                                                      performance of their own data streams
                                                                      Passive – introduce traffic only in the form of monitoring data and requests
                                                                      for monitoring
                                                                                              Tasks Involved
                                                                       Develop a monitor activation mechanism
                                                                      Develop monitor software and hardware
                                                                      Develop data collection and display capabilities
                                                                      Deploy monitors
                                                                      Work with applications


          Impact and Connections                                                  Milestones/Dates/Status
 IMPACT:                                                                   Monitor Daemon                                   Year
      Build a monitoring infrastructure that will aid in debugging        - Design base passive monitor daemon                 1
       of distributed application communication and support both           - Activation mechanism integration                   1
                                                                          - Improvements to network drivers                     1
       active and passive monitoring                                      - Improvements and enhancements to sensor mechanism 2 & 3

 CONNECTIONS:                                                        •   Activation Mechanisms
                                                                         - Design basic activation mechanism                        1
     SciDAC: Net 100, DOE Science Grid, Astrophysics,                   - Develop and deploy full activation capabilities         2&3
       Bandwidth Estimation, Data Grid, INCITE, Net100                 Results Handling Infrastructure
     Base:Network Monitoring, Data Grid, Transport Protocols            - TCP dump viewing capabilities                            1
 URL:                                                                   - Develop improved data viewing capabilities              2&3
                                                                       Deployment of Monitors
     www-itg.lbl.gov/Net-Mon/Self-Config.html                          - Deployment to initial ESnet sites (gig-E)                1–3
                                                                        - Work with applications                                   2&3
                                                                        - Additional ESnet sites                                   2&3


                                 MICS Program Manager: Thomas Ndousse                                                              Date Prepared: 1/7/02
                                      High-Performance Transport Protocols
                  PI: Wu-chun (Wu) Feng, Los Alamos National Laboratory and The Ohio State University

         High-Performance Network Research -                                                                       Base Project

Goal: To significantly improve network                                                  The Novel Ideas
performance in support of all computing                                  Dynamic Right-Sizing: TCP Flow-Control Adaptation
environments, particularly grids and NGI.                                 for Grids & the Next-Generation Internet
• TCP/IP  Make the network fast but TCP friendly.                           Automatically enhance network performance over the
   • Eliminate TCP’s flow-control bottleneck                                  WAN by as much as an order of magnitude while
     by automatically tuning buffer sizes.                                    abiding by TCP semantics.
• RAPID  Make the network more adaptable.                               RAPID: Rate-Adjusting Protocol for Internet Delivery
   • Smooth QoS support over a best-effort
     network.                                                                Provide smoother QoS support over the best-effort
   • User-settable reliability, providing a spectrum                          Internet for grids and NGI while minimizing the need for
     of QoS from unreliable UDP to reliable TCP.                              widespread deployment of DiffServ or IntServ.


          Impact and Connections                                                 Milestones/Dates/Status
 IMPACT.                                                                                                               Mon Yr DONE
      Dynamic Right-Sizing                                             Simulation: Flow-Control Adaptation with Dynamic Right-Sizing
         - Auto-tuned, order-of-magnitude increase in throughput.          -Protocol Analysis & Design (ns-2)           12/01     12/01
         - Vendor adoption, e.g., IRIX, Linux (still in the works)         -Protocol Testing & Evaluation (rudimentary) 03/02 beta testing
         - Potential integration into GridFTP, Web100, Net100.          Implementation: Flow-Control Adaptation with Dynamic Right-Sizing
      RAPID                                                               -Kernel Space, Linux 2.4.x                   07/02 beta testing
         - Sliding reliability semantics may result in adoption of         -User Space, drsFTP                          01/03 alpha testing
           RAPID by LANL large-data visualization team.                    -Protocol Testing & Evaluation (rudimentary) 03/03
 CONNECTIONS.                                                             -Potential Integration with GridFTP          04/03
      Dynamic Right-Sizing: Web100, Net100, DOE Science Grid,             -Deployment (kernel- & user space)           07/03
       Particle Physics Data Grid, Earth System Grid II,                Simulation: RAPID
      RAPID: The LANL large-data visualization team, previously           -Effect of packet spacing                    03/02 preliminaries
       sponsored by the DOE NGI Corridor One project. Others?              -Definition of API to middleware             03/02 preliminaries
                                                                           -Sliding reliablity                          07/03
                                                MICS Program Manager: Thomas Ndousse-Fetter                               January 16, 2002
                                                         IQ-ECho
              PIs: Schwan, Ahamad, Eisenhauer, Yalamanchili -- Georgia Institute of Technology

     High-Performance Network Research                                                                              Base Project

       IQ-ECho – Interactive Quality of Service                               IQ-ECho Novel Ideas
       Across Heterogeneous Hardware/Software
                                                                    • integrated QoS management through quality attributes
  represent information flows as event streams in                  • dynamic code generation relocates application-level
   event-based IQ-ECho middleware                                     functionality to the most appropriate location
  use dynamic code generation to migrate
                                                                    • configurable protocols and kernel-level monitoring provide the
   application-level filtering/ data processing to
                                                                      system-level support required for online quality management
   appropriate network locations
  use network-level feedback to drive application-                 • vertical programming allows extending platforms while
   level quality of service adaptations.                              programming applications

      http://www.cc.gatech.edu/systems/projects/IQECho


         Impact and Connections                                                 Milestones/Dates/Status
 IQ-ECho IMPACT.                                                       Year 1                                            Mon Yr       DONE
    – enable network-aware adaptable applications                     • performance attributes in ECho middleware         4/02
    – cross-layer information exchanges will make effective           • select and implement sample application           6/02
      runtime tradeoffs in quality vs. performance across the         • create instrumentation for performance attributes 8/02
      protocol, middleware, and application levels                       Year 2
    – enable the creation of efficient and adaptable Grid data        • evaluate and tune middleware                      3/03
      services                                                        • enable application for adaptation                 3/03
 CONNECTIONS:                                                        • extend/create configurable network protocols      6/03
    – Remote visualization (Supernova Visualization), source-         Year 3
      based filtering (Oakridge), program monitoring and steering     • integrate ECho-IQ with access grid software       3/04
    – Extensible cluster platforms (NSF, DOE)                         • demonstrate benefits in access grid environment 6/04
    – Remote sensing, monitoring, and security (DARPA, NSF)



                                MICS Program Manager: Thomas Ndousse                                                 Date Prepared: 1/10/02
                                                          PingER
                                                   PIs: Les Cottrell SLAC

      High-Performance Network Research                                                                                  Base Project
         PingER: Active End-to-end performance                                       PingER novel ideas
          monitoring for the Research and Education                   Low impact network performance measurements to most of the
                         communities                                 Internet connected world providing delays, loss and connectivity
   Tasks:                                                            information over long time periods
   -develop/deploy simple, robust ssh based active                    Network AND application high throughput performance
   end-to-end measurement and management                             measurements allowing comparisons, identification of bottlenecks
   infrastructure                                                     Continuous, robust, measurement, analysis and web based
    -develop analysis/reporting tools                                reporting of results available world wide
    -integrate new application and network                            Simple infrastructure enabling rapid deployment, locating within an
   measurement tools into the infrastructure                         application host, and local site management to avoid security issues
    -compare & validate various tools, and determine
   regions of applicability
      www-iepm.slac.stanford.edu
                                                                                 Milestones/Dates/Status
        Impact and Connections                                          Infrastructure development             Mon/Yr     DONE
                                                                       - develop simple window tuning tool      08/01      08/01
 IMPACT:                                                              - initial infrastructure developed       12/01      12/01
      increase network and Grid application bulk throughput           - infrastructure installed at one site   01/02      01/02
       over high delay, bandwidth networks (like DOE’s ESnet)          - improve and extend infrastructure      06/02
                                                                       - deploy at 2nd site                     08/02
      provide trouble shooting information for networkers and
                                                                       - evaluate GIMI/DMF alternatives         10/02
       users by identifying the onset and magnitude of                 - extend deployment to PPDG sites        03/03
       performance changes, and whether they appear in the           •  Develop analysis/reporting tools
       application or the network                                      - first version for standard apps        02/02
      provide network performance data base, analysis and            Integrate new apps &net tools
       navigateable reports from active monitoring                     - GridFTP and demo                       05/05
 CONNECTIONS:                                                         - INCITE tools                           08/02
                                                                       - BW measure tools (e.g. pathload)       01/03
      SciDAC: High Energy Nuclear Physics, Bandwidth                • Compare & validate tools
       Estimation, Data Grid, INCITE                                   - GridFTP                                09/02
      Base:Network Monitoring, Data Grid, Transport Protocols         - BW tools                               04/03
                                MICS Program Manager: Thomas Ndousse                                                          Date Prepared: 1/7/02
       Stability Modeling and Control of Transport Protocols for High-Speed Data Grids
                                               Nageswara S. Rao, Oak Ridge National Laboratory

        High-Performance Network Research                                                                              Base Project

 Understand and Control the End-to-End                                                   The Novel Ideas
 Transport Dynamics of High-Speed Grids                                Detailed analysis of transport dynamics using non-linear control and
 • Detailed analysis of transport processes                            chaos theory – showed that TCP generates “complicated” phase
      • rigorous treatment using non-linear control                    space attractors
         and chaos theory                                              Developed the concept of grid network instruments to perform
                                                                       measurement and traffic engineering using light-weight in-situ modules
 • Develop provably effective transport                                – analytically showed their performance optimality
 methods for:                                                          Novel transport control methods for end-to-end control for
         •high throughput, and                                                  - high throughput using concurrent window and graded control
         •end-to-end dynamics control                                           - controlled dynamics using multiple throttle methods
 • Implement and test on grid environments


          Impact and Connections                                                 Milestones/Dates/Status
 IMPACT.                                                               Detailed rigorous analysis:
      Provides controlled end-to-end dynamics for grids over wide-       - attractor analysis                           Feb 02/Feb 03
       area networks – significant step beyond state-of-the art           - conditions of chaos                          Apr 02/Apr03
      Fundamentally new classes of transport methods based on          Grid network instrumentation design:
       sound analysis and experimentation – inexpensive and easy          - sufficiency proofs of measurements           Mar 02/ Mar03
       to use                                                             - detailed module design                       June 02
      Provides the needed quality of service for control over wide-    Proof of concept implementations:
       area networks for data and instrument grids                        - high throughput                              July 02
 CONNECTIONS:                                                            - bounded higher order delay moments           Aug 02/Sept 03
      Net100 project: will use the proposed instruments and will       Application and testing:
       provide certain measurement modules                                - identification of representative problem      Feb 03
      Terascale Supernova Initiative can significantly benefit from      - performance study                             Sept 03
       the proposed control methods – we are in communication

                                   MICS Program Manager: Thomas D. Ndousse                                             Date Prepared: 01/09/02
                            Pushing the Network Simulation Envelope
                                       W. R. Wing - Oak Ridge National Laboratory

     High-Performance Network Research                                                                                 Base Project
                                                                                        SSFnet Novel Ideas
      SSFnet - Creating a Terascale network
   simulator that can model SciDAC applications                      • SSFnet will be the first network simulator with verifiable instrumentation
                                                                       - We plan to include (not model) the Net100/Web100 MIB
  Tasks:                                                               - Net100/Web100 MIB data will be accumulated for direct comparison
  - Verify SM SSFnet on candidate architectures
                                                                     • SSFnet will be the first production quality Distributed Memory simulator
  - Develop initial DM version of SSFnet
                                                                       - Domain Modeling Language will automate decomposition
  - Develop and verify instrumentation
  - Develop application-level IDE                                    • SSFnet will be the first simulator able to tackle SciDAC-scale problems
  - Distribute to DOE network research community
  - Develop 2nd-Gen DM scheduler and DML




         Impact and Connections                                                  Milestones/Dates/Status
 IMPACT: SSFnet will be the first network simulator able to:
      Fully model SciDAC Terascale applications                         Proposed Milestone                     Proposed Date       Actual Date
      Allow SciDAC developers to tune their applications to
        evolving mixed-technology network environments                 Verify Shared-mem architectures
                                                                        - IBM, Compaq, Solaris        Q1 - FY02 Complete
      Allow testing/confirmation of future SciDAC-developed           Develop initial DM scheduler    Q3 - FY02
        network protocols                                              Develop MIB instrumentation      Q4 - FY02
 CONNECTIONS: A key element of SSFnet’s verifiability is our          Develop application-level IDE   Q2 - FY03
                                                                       Develop 2nd-Gen DML-based Scheduler Q4 - FY03
  plan to directly incorporate the Net100/Web100 MIB in the            Distribute to DOE community      Q4 - FY03
  simulator. Comparison of real-life MIB measurements with the
  SSF-instrumented MIB will provide confirmation of SSFnet
  simulation fidelity. However, this does require deployment of at
  least some SciDAC applications on Web100/Net100 platforms



                                 MICS Program Manager: T. Ndousse                                                        Date Prepared 01/08/ 02
                                      High-Performance Transport Protocols
                  PI: Wu-chun (Wu) Feng, Los Alamos National Laboratory and The Ohio State University

         High-Performance Network Research                                                                         Base Project

Goal: To significantly improve network                                                  The Novel Ideas
performance in support of all computing                                  Dynamic Right-Sizing: TCP Flow-Control Adaptation
environments, particularly grids and NGI.                                 for Grids & the Next-Generation Internet
• TCP/IP  Make the network fast but TCP friendly.                           Automatically enhance network performance over the
   • Eliminate TCP’s flow-control bottleneck                                  WAN by as much as an order of magnitude while
     by automatically tuning buffer sizes.                                    abiding by TCP semantics.
• RAPID  Make the network more adaptable.                               RAPID: Rate-Adjusting Protocol for Internet Delivery
   • Smooth QoS support over a best-effort
     network.                                                                Provide smoother QoS support over the best-effort
   • User-settable reliability, providing a spectrum                          Internet for grids and NGI while minimizing the need for
     of QoS from unreliable UDP to reliable TCP.                              widespread deployment of DiffServ or IntServ.


          Impact and Connections                                                 Milestones/Dates/Status
 IMPACT.                                                                                                               Mon Yr DONE
      Dynamic Right-Sizing                                             Simulation: Flow-Control Adaptation with Dynamic Right-Sizing
         - Auto-tuned, order-of-magnitude increase in throughput.          -Protocol Analysis & Design (ns-2)           12/01     12/01
         - Vendor adoption, e.g., IRIX, Linux (still in the works)         -Protocol Testing & Evaluation (rudimentary) 03/02 beta testing
         - Potential integration into GridFTP, Web100, Net100.          Implementation: Flow-Control Adaptation with Dynamic Right-Sizing
      RAPID                                                               -Kernel Space, Linux 2.4.x                   07/02 beta testing
         - Sliding reliability semantics may result in adoption of         -User Space, drsFTP                          01/03 alpha testing
           RAPID by LANL large-data visualization team.                    -Protocol Testing & Evaluation (rudimentary) 03/03
 CONNECTIONS.                                                             -Potential Integration with GridFTP          04/03
      Dynamic Right-Sizing: Web100, Net100, DOE Science Grid,             -Deployment (kernel- & user space)           07/03
       Particle Physics Data Grid, Earth System Grid II,                Simulation: RAPID
      RAPID: The LANL large-data visualization team, previously           -Effect of packet spacing                    03/02 preliminaries
       sponsored by the DOE NGI Corridor One project. Others?              -Definition of API to middleware             03/02 preliminaries
                                                                           -Sliding reliability                         07/03
                                                MICS Program Manager: Thomas Ndousse-Fetter                               January 16, 2002

				
DOCUMENT INFO