CCIRN Measurements Working Group meeting December 12, 1997 Washington, DC Draft Notes I. Participants Guy Almes, I2 Julio Andrade, DynCorp Bill St. Arnaud, CANARIE Kilnam Chon, KAIST John Coulter, CANARIE Shigeki Goto, Waseda U Rene Hatem, CANARIE Sandra Jaque, REUNA Hyungsuk Kim, KT David Martin, Fermilab/DoE Kevin Meynell, TERENA Toh Ee Ming, NUS-IRDU Singapore Masaaki Nateshima, NTT Niels den Otter, SURFnet Roberto Sabatino, DANTE Lai Zit Seng, NUS-IRDU Singapore Byung-Cheol Shin, KAIST George Strawn, NSF Shuichi Tashiro, ETL II. Actions Items * Professor Goto will call a working group meeting in the first week of March at the APAN meeting. * Begin to work on a proposal for collaboration on International Research Networks * At ISMA(scheduled for early April) the measurement group will put together a common set of tools. * Guy Almes to bring a proposal of the IP Performance for review. III. Proceedings A. Global Traffic Measurement Guy Almes, I2, gave a presentation of I2's IP Performance work in the measurements area. Their findings show that the topology is becoming increasingly complex, loads are growing faster then capacity, and the relationship among networks is increasingly competitive. Their objective is to facilitate the complex relations among networks. They have been working on a collaboration of advanced CSG surveyor statistics. The infrastructure will be built with a series of measurement machines. There will be a central database with a web server to disseminate results. The measurement machines are constantly recording date and uploading it to the web server for access. At this time there are 12 operational sites which are measuring delay and loss of packets. They will be moving towards connecting 23 more labs, including the GigaPops of I2. Guy presented a sample of the data that is retrievable from the web site on overhead slides. This data shows the amount of data that is lost and delayed at certain periods of time. David Martin, DoE ESNet, presented on the ESNet Ping ER effort. This effort began by measuring all the routes a packet took in transit. Through this they defined a link as the flow of traffic from one end point to the other. ESNet is using ICMP Ping packets to measure availability and delay. Analysis sites are setup to provide historical reporting, unlike I2 whose data is updated and available immediately. The collecting sites run PERL Script and CGI making date available via HTTP, while maintaining a list of remote sites. The analysis sites retrieve data from the collecting sites, polling them sequentially for the last 48 hours. Each collection site is stored in a single SAS dataset for each month. ESNet has found ICMP to prove itself as a good measurement of network performance. The effort to keep these sites running is significant. All users want a single measure of link quality, using reasonable parameters. The historical data is useful, but ESNet would like to have up to the minute date on select sites in the future. They are also looking at better grouping and presentation tools to display data collected. This presentation led to a discussion on how to use a common tool for measurement internationally. The main focus in measurements is cost sharing, for example, connecting at the STAR TAP. It was suggested, that although proprietary, that Net Flow is being used by some of the participants in the working group and it has proved to be very useful. Kilnam Chon asked to see a proposal of I2's IP Performance to review and possibly find some useful tools in measuring the international connectivity. The group agrees that the largest barrier to moving forward with connecting internationally is the session initiating. At this time their is no way to tell who initiates the connections across these lines. The question becomes does more difficult, for example, does the US want the other countries to have this information, therefore implying that the US is initiating the connection. Or, is the information readily available and the other countries are looking to obtain this information, thus making them the initiating party. The group ended the discussion of a review of the action items, listed above. Julio D. Andrade DynCorp I&ET jandrade@snap.org