January 25, 2004 from 2:00 - 5:00 pm

SPEAKERS
Eric Boyd, Internet2 [PPT]
Jeff Boote, Internet2
Tanya Brethour, NLANR

Challenged by members of the E2E TAG, the two teams have created interoperable measurement frameworks, using a measurement schema developed by the GGF Network Measurement Working Group (NMWG). The demo illustrates the ability to do partial path analysis of the University of Hawaii <-> Hawaii POP <-> Abilene Backbone <-> University of Illinois Urbana-Champaign end-to-end path.

Eric Boyd opened the meeting with an oaverview of the E2E piPEs effort and described what parts of the architecture are actually complete and which ones are under development.. He described the original scope of the project and how it has increased to incloude more ingetral steps along the way - focusing on the need for measurement boxes at more frequent intervals.

Tanya Brethour presend and overview of the Advisor project - she described the Performance Data Collectors and the bundles of tests that are available. She noted that the data archiver makes the data available to any user. She described the metrics that are being used by the analysis engine; the new features will still contact the user if it detects any of the problems it recognizes. The Advisor GUI includes both an expert GUI (all metrics displayed), a simple analysis guide (simple text advice) and a map GUI (which is a more visual, graphic version of the data shown in the expert GUI). She showed the expert GUI for the alpha release; she suggested that folks try to download the current release and report on it. Tanya expects the beta version to be ready in late April.

Jeff Boote described the measurment boxes on the Abilene Network - that there are regular tests going on. For the purposes of the demo, they extended this out to the Hawaii PoP so that they could extend the measurements out to the "desktop", running regular tests between Denver to/from Hawaii and Seattle to/from Hawaii. Currently, they are experiencing delays for the past two hours from Seattle to Hawaii. There's an OC3 in between here so there is some congestion. OWAMP describes the one-way delays (latency) on this connection; the demo is also running Iperf for throughouput measurements - over past 24 hours, the throughput has been steady.

To do a full mesh of nodes (11) with scheduled Iperf tests, Internet2 had to create the BWCTL (Bandwidth Control) tool to manage the time slots. It checks with both hosts and allocates the next available time slot. This was dones so that people can schedule on-demand Iperf tests to the middle of the network.

Tanya described the demo - she had the Expert GUI running on her laptop, a PDC on the Indianapolis Abilene node, a workstation at her home location UIUC and another one at the University of Hawaii. The data is being fed into the one database (group) with a web service setup. As the "end user", Tanya contacted the web servcie to ask for their Iperf and OWAMP data.

Eric Boyd described an end-to-end performance problem that they ran into when setting up the tests; at firts, the tests of the Hawaii PDC was getting only 200 Kbps from Seattle to Hawaii. Using the E2E piPEs analsysis portion (which includes the NDT), they located a known "flaky" box. Once they pointed this out to the system administrator for that region, they were plugged into a different box and all the problems went away. This proved the success of the E2E piPEs framework for locating localize problems.

Several questions were raised; among them was a question on how the E2E piPEs framework was related to the old Surveyor project. Jeff reported that Surveyor was only round-trip pings and was very regimented in terms of scheduling. The E2E piPEs effort is attempting to make routine data available on a regular basis and still allow end-users to schedule on-demand tests within a regional timeframe.

Another question dealt with whether the teams are dealing with international efforts; Eric reported that E2E piPEs is working closely with GEANT and APAN at this time. The HENP-funded MonALISA project, which runs over several networks, is being coordinated with the Internet2 effort.