problems we need statistics to solve


need

  • trace-driven experiments

    why

  • cache management
  • optimize queueing
  • congestion and scaling dynamics
  • aggregate transport behavior

    measurements needed

  • full header traces

    where

  • small number of high aggregation sample points
    near NAPs
    portable
     kc@caida.org, 20 feb 1996

    problems we need statistics to solve

    need

  • specifications for ISPs to give routers vendors

    why

  • router benchmarking

    measurements needed

  • flow counts/parameters
  • traffic matrices

    where

  • NAPs
  • backbone core points
  • corporate campuses
  • large customers
  • private ISP interconnects


    kc@caida.org, 20 feb 1996

    problems we need statistics to solve

    need

  • specifications for users to give ISPs

    why

  • service quality assessements
  • ISP shopping

    measurements needed

    multi-function transponder testing platforms

    where

  • bridged at strategic interconnects
  • uncongested customers
    kc@caida.org, 20 feb 1996

    problems we need statistics to solve

    need

  • routing system stability

    why

  • RAS

    measurements needed

  • route (BGP) logging and archiving
  • caching architecture

    where

  • bridged at routing exchange points
  • large customers

    kc@caida.org, 20 feb 1996



    problems we need statistics to solve
    need whymeasurements neededwhere
    trace-driven experiments
  • cache management
  • optimize queueing
  • congestion and scaling dynamics
  • aggregate transport behavior
  • full header traces
    		• few high aggregation points (near NAPs)
    specifications for ISPs to give routers vendors router benchmarking
  • flow counts/parameters
  • traffic matrices
  • NAPs
  • backbone core points
  • corporate campuses
  • large customers
  • private interconnects
    • specifications for users to give ISPs
  • service quality assessements
  • ISP shopping
    		• multi-function transponder testing platforms
  • Mathis' treno
  • Paxson's probe daemon
  • bridged at strategic interconnects
  • uncongested customers
    • routing system stability RAS
  • route (BGP) logging archiving
  • caching
    		• route servers


    general needs

    • test points and colocation space
    • ongoing dialogue between networking researchers and operational data collectors
    • transponder platform most useful if standardized
    • network service procurement language for customers
    • protocol and application developers cognizant of operating environment

    kc@caida.org, 20 feb 1996