Draft Notes

I.  Participants
Michael Behringer, DANTE
Suzanne Burgess, DynCorp/FNC
Woohyong Choi, KAIST
Kilnam Chon, APAN/KAIST
John Dyer, TERENA
Jan Gruntorad, CESNET
Lawrence Law, HARNET
Kevin Meynell, TERENA
Tan Tin Wee, National University of Singapore
Forencio Utreras, REUNA
Karel Vietsch, TERENA

II.  Action Items
- Distribute materials on the Common Solution Group framework for collecting traffic performance measurements. (Burgess)
-  Propose a meeting after December IETF (in WDC) to identify and coordinate current measurement efforts.  Explore funding options for this meeting. (Burgess)
-  A working group chair needs to be appointed. (CCIRN)

III.  Proceedings:
A.  Summary of ISMA 97
Kevin Meynell(TERENA) and Bill Manning (USC/ISI) provided the group with some of the main conclusions from the Internet Statistics and Metrics Analysis (ISMA) workshop.  The workshop was an invitational meeting for individuals involved in developing or deploying Internet traffic measurement or analysis tools. Sixty-five (65) people attended, representing Internet service providers (ISP), the research and education (R&E)  community, vendors, and end-users.

Workshop participants concluded that current measuring tools can measure certain activities on the Internet, but there still is no comprehensive tool to measure and report on  Internet activity.  Other conclusions were that measurement does not equate management and there is no concrete definition of what exactly Internet "performance" is. Obstacles to collecting stats include the feeling that ISPs treat their statistics as corporate or proprietary information still exists.  Furthermore, the variety of protocols used in the Internet, lack of common trouble tickets, and the additional training required in gathering and analyzing stats/metrics were also discussed at ISMA.  Many are revisiting the IETF benchmarking effort so that an agreed-to standard can be used to tell ISPs what information is required and why.

Steps being taken include the use of GPS for date stamps for resolutions and the agreement that GMT is the standard for that date stamp.  The vBNS has found the OC3MON (described at http://www.nlanr.net/NA/Oc3mon/) useful in its measurement activities and looking to deploy it elsewhere.

The report on the conference is found at http://www.nlanr.net/ISMA/isma97_report.html

B.  TF-European Traffic Measurement (ETM)
Kevin Maynell (TERENA) briefed the group on the TERENA Traffic Measurement taskforce which began last November.  Influencing factors identified for creating this task force were the increasing traffic and the desire to better manage resources.  Furthermore, consumers are beginning to ask about Quality of Service from ISPs so that when they explore different Internet service providers they can evaluate the offerings.

At the first meeting, the role of a European task force on stats/metrics was discussed because they did not  want to replicate current efforts.  It was decided that while some overlap was inevitable, the nature of the Internet in Europe was unique enough to offer a different perspective.  Near term, the group plans to circulate a survey within three months asking what statistics are currently gathered and what networks would like to measure.  The group will also examine the list of tools on the CAIDA site and evaluate measurement tools to develop a recommended tool kit

Information on the task force is available at http://www.terena.nl/task-forces/tf-etm/ and to join its mail list send a message to tech-staff@terena.nl.

Meynell reported that a RIPE-NCC project just started and is funded by RIPE.  It will publish inter-AS data which several ISPs were demanding.  Others were concerned that the results are scientifically defensible.  Another measurement activity is just started in  Israel is collecting incoming traffic data on well-known TCP/UDP ports.

C.  TEN-34
Behringer briefed the group on the measurement plans for the TEN-34 network.  The plan is to run additional equipment inside the network with one workstation per POP.   Discussions are underway on the workstation specifics and the specific performance measurement tools to be used on the network.  Currently the priority is to setup Netflow.  After the initial stage, ATM measurements could be monitored.  A brief description of the TEN-34 implementation plan is available at http://www.dante.net/ten-34/impl.html

D.  APAN
A measurements working group within APAN was just formed and this group is concerned with ensuring as much throughput as much as possible.  The group is planning to explore bandwidth delay and latency.  More information on this effort can be found on the APAN web site (http://www.apan.net).

E.  Discussion
The group agreed that dependence on a vendor collection method is a recipe for disaster and networks (TEN-34 and APAN) need to verify them these methods from an independent source.  Most would prefer to use OC3MON, but that is not always possible.

There are currently many efforts - NLANR, Common Solutions Group (http://www.advanced.org/csg-ippm/), National Internet Measurement Infrastructure (NIMI) (http://www.psc.edu/networking/nimi/welcome.html), Internet Protocal Performance Metrics (http://www.advanced.org/IPPM/), Internet Performance and Measurement Analysis (www.merit.net/ipma/), CAIDA (http://www.caida.org - are underway in the area of IP measurement, but it was felt that there is a need to disseminate information from these group more widely and report on the use of different tools already developed.  (See above action for a possible workshop after the December IETF).

The group also wanted to hear more about efforts in the area of ATM measurement tools.