NSF Workshop on Internet Statistics Measurement and Analysis

San Diego Supercomputer Center, 19-20 February 1996.

List of questions related to each discussion topic

Topic 1: Workload Characterization:

• What are the statistical characteristics of measured traffic: Long-range dependence, heavy-tailed distributions?
• What assumptions are made now based on traditional models? Are these assumptions significantly misleading?
• Is network dimensioning practiced as a black art?
• How can providers get useful analysis from researchers and academics?
• How should funding agencies direct funding in useful directions, and maximize cooperation between providers and researchers?

Topic 2: Measurement Infrastructure:

• What measurement capabilities exist now in switches and routers?
• What traffic trace data exists? How widely available is it?
• What traffic analyses have been published? How useful are they?
• Are ISPs cooperating with measurement efforts?
• Are ISPs conducting their own measurements?
• What are standards requirements for routers and switches?
• What's happening in the IETF Internet Provider Performance Metrics (IPPM) group?
• What's happening in the IETF Benchmarking working group (BMWG)?
• What other related standards activity exists?
• Large-scale topology engineering; using imperfect measurements to do realistic and reasonable traffic management.
• Who is promoting independent assessment of ISP performance?

Topic 3: Flow Characterization and Control:

• How is the TCP/IP flow duration distribution changing? What impact?
• How is the TCP/IP route locality distribution changing? What impact?
• Effect of web (short-lived) traffic and MBONE (heavy) traffic.
• Interaction among providers in response to new applications.
• Effect of high aggregation, large number of small flows.
• What is happening in the trans-Atlantic links w.r.t. TCP connection startup and efficiency? (Lothberg, Claffy, Paxson may know.)

Topic 4: Internet Scaling:

• What changes technically as network gets very large?
• Is better reliability the necessary consequence of the Internet getting bigger?
• How do ISPs control their networks? How would telephone companies do it under their business/technical models?
• Which differences are implied by scale, which by local precedent, culture/history?

Topic 5: Quality of Service and More Complex Pricing:

• Does a more complex QoS model imply a more complex pricing model?
• TCP/IP congestion control operates at the edges. How does provider take responsibility for quality/reliability of its network?
• Link sharing concept; prospects for practice?
• Issues of measurement related to billing.

Topic 6: Privacy and Security:

• What is current practice and capability in deployed technology?
• What interesting tools are available and used?
• Examples of recent security-comprimising events.
• What are the outstanding hard problems?
• What information is sensitive for providers?
• What are legal restrictions?
• How do regulations of phone network and internet differ?

Topic 7: Cooperation:

• The continuing role of the Federal Government as Internet shepherd.
• Market competition and competetitors: how do they cooperate?
• One-on-one relationships vs. activities en mass (such as this workshop).
• Cooperation between providers and researchers.
• Cooperation and interaction across national boundaries; other governments?
• Creation of further focused workshops and meetings.
• What other mechanisms?
• Can we use technical input here to generate a list of the most important technical problems for NSF to pursue.
• Establish concrete action items to take home.
• Outline workshop report; identify value.