Federal Collaboration in Internet Statistics Collection

  Slides presented at the NSF sponsored workshop on Internet
  statistics measurement and analysis held at SDSC 19-20 Feb 1996.

  Phillip Dykstra
  Federal Networking Council
  Eng. and Ops. Working Group, co-chair

  Army Research Laboratory
  Attn: AMSRL-SC-CC
  APG, MD 21005-5067
  Voice: 410-278-8608
  FAX: 410-278-9199
  Email: phil@arl.mil


Federal Collaboration
 * The Federal networks and government has *some* influence
   over the internet and can help provide focus
 * We would like industry to do it
 * We have a traditional role of collecting and publishing
   data of national interest

The FNCAC Asked for...
 * Requirements for service models
 * Accounting and pricing architectures with incentives to
   place a value on resource consumption
 * Broad integration of security and privacy measures
 * Cost/benefit tradeoffs of "doing" or "not doing" things

The Bitstream is Important
 * Bandwidth planning
 * Topology planning
 * Problem identification (failures, routes, etc.)
 * Service planning
 * Types of flows
   e.g. emergence of many small transactions (HTTP) and
   long sustained flows (MBONE)
 * Deployment of security

Engineering Isn't Everything
 * Not everything is in the bitstream
   # of routes and route stability
   # of NSP's
   # of users
   User satisfaction
     Demographics (Nielson families/ratings)
     Health/Security (cancer statistics)
 * Wise investment requires lots of data
 * Leading Internet Indicators (LII)?
     New starts?
     Consumer confidence?
     Excess capacity (always = 0?)

Issues
 * How to measure "the net" when no one owns it?
   What is the federal role and relations?
 * Who is responsible for collecting and publishing statistics?
   The government, a private-sector org., etc.?
 * How can privacy be ensured?
   Can we looks beyond packet headers?

See also a paper at http://www.fnc.gov/metrics.html