RIPE 61

Measurement, Analysis and Tools Working Group
RIPE 61, Rome
Thursday 18 November 2010, 16:00-17:30

Status: Final

Scribe: Rene Wilhelm, RIPE NCC
Chat: Chris Buckridge, RIPE NCC

A. Administrative Matters

Co-chair Ian Meikle welcomed attendees and opened the first session of new MAT Working Group.

B. Traffic Identification Engine - Alberto Dainotti, University of Naples

The presentation is available at:
http://ripe61.ripe.net/presentations/266-tie-ripe-v5.pdf

Alberto presented the research done in the University of Naples on traffic classification. They would appreciate feedback from the operator community. What are the needs of and technical problems encountered by ISPs in the field of traffic classification? The group is also interested in collaborations with service providers.

Steve Nash from ARBOR Networks stated he was interested to talk to Alberto about the plug in concept.

C. Neustar Anycast Statistics- Edward Lewis, Neustar

The presentation is available at:
http://ripe61.ripe.net/presentations/351-RIPE61MATWG-Lewis.pdf

Edward Lewis talked about the DNS statistics gathering system Neustar have in place for about a year now.

Daniel Karrenberg from the RIPE NCC asked if the software was in a proprietary package; if there was a chance to get insight into the specifications of how Neustar return things so others could learn and build something themselves.

Edward Lewis replied there is no deliverable software. It's not just a package; it's built into the core. Providing an overview of what the product does would be more of a sales presentation.

D. Information Services Update - Sean McAvoy, RIPE NCC

The presentation is available at:
http://ripe61.ripe.net/presentations/352-Information_Services_Update_RIPE_61.pdf

Sean McAvoy gave an update on developments in RIPE NCC Information Services. An improved UI for the DNSMON service has been released as public beta.

E. IPv6 Network Measurement - Owen DeLong, Hurricane Electric

The presentation is available at:
http://ripe61.ripe.net/presentations/353-RIPE61_Measurement.pdf

Owen DeLong presented some statistics on IPv6. Countries on v6, growth in routing table, types of network traffic and signs of a first (useless) attack in IPv6.

F. Visualizing geolocated Internet measurements - Claudio Squarcella, CAIDA

The presentation is available at:
http://ripe61.ripe.net/presentations/201-preso.pdf

Claudio presented some new techniques to visualize Internet measurements based on geographical location.

Klaus Darilion, from nic.at, enquired about the software used for these visualisations. Claudio explained he used Google web toolkit, the Raphael JavaScript library (for graphics) and Java code translated into Javascript.

G. Atlas Technical Details - Robert Kisteleki, RIPE NCC

The presentation is available at:
http://ripe61.ripe.net/presentations/269-20101118-RIPE61-MAT-Robert.pdf

Robert presented the technical aspects of the Atlas project. Design principles as well as implementation details.

Wolfgang Tremmel, via the chat, suggested to provide the firmware sources for download in order to prevent hacking.

Daniel Karrenberg replied RIPE NCC is not going to provide the source code any time soon. Making RIPE Atlas open source at this moment, while it is just starting would put its success at risk. Someone could take the idea, the design and commercialise it into a competing product. Once the project has established itself, has sponsors and there is more experience, RIPE NCC will be very open to input and requirements from sponsors and the membership regarding these issues.

Jelte Jansen from ISC asked how probe re-boots, which could be needed to recover from memory exhaustion, are going to be handled. Will users receive reboot request via e-mail? Robert explained that a watchdog timer handles this. The probe re-boots automatically, is back in business within a minute or so.

Ruben van Staveren asked, via the chat, how fast RIPE NCC could provide updates in case of misbehaviour or security issues. Robert replied that the RIPE NCC could disable probes; if a probe is found misbehaving it can be taken off the measurement net immediately. Deploying a new firmware can be done within days, once it is ready. The actual development time for new firmware depends on the amount of changes needed. Daniel Karrenberg added that RIPE NCC does thorough testing before distributing new firmware to the production network.

Richard Barnes from BBN commended the RIPE NCC on the design of the system. He stresses the major value of the Atlas network is in the large volume of data it produces. Next to graphical user interfaces and visualisations, he would like to see a programmatic interface, an API to access the raw data.

H. Measurement Framework Proposal and Discussion - Daniel Karrenberg,

RIPE NCC and the WG Co-Chairs

There was a lively discussion on the proposal by the WG chairs of having a single platform that not only shows results from RIPE NCC tools, but also from other organisations, institutions who would be happy to share their data. Daniel added to that his vision of building a measurement community by providing the means to annotate, share and discuss results.
This would be the place to go whenever something comes up, an anomaly or strange behaviour of the Internet.

The attendees did not oppose the proposal, it was suggested to try it out and see how it evolves.

Co-chair Richard Barnes reminded the audience of the mat-wg _at_ ripe _dot_ net mailing list. He added that it has seen little traffic since the last RIPE meeting, but this will change as things get started. The list will be the venue for collecting community feedback on these topics. Also, RIPE NCC will report more often to the list. People are encouraged to subscribe to the mailing list and participate in the discussion.

Richard thanked the audience for attending and closed the session.