Measurement, Analysis and Tools Working Group
RIPE 62, Amsterdam
5 May 2011, 16:00-17:30
Revision Number: 1
Co-chairs: Richard Barnes, Christian Kaufmann, Ian Meikle
Scribe: Devin Bayer, RIPE NCC
Chat: Vasco Asturiano, RIPE NCC
A. Co-chairs' Intro
Christian Kaufmann, WG co-chair, said they were seeking speakers for the upcoming ENOG conference.
Christian’s presentation is available at:
B. RIPE Atlas update - Robert Kisteleki, RIPE NCC
Robert’s presentation is available for download at:
Florian Obser, Wyneken & Schaefer, commented that the probes couldn’t work with only IPv6 connectivity.
Robert replied that they are working around the issues.
Mohsen Souissi, AFNIC, asked about the criteria for the planned custom DNS measurements, especially in regards to hardware resources.
Robert suggested discussing the possibilities and the Chair said it would be a good topic for the Measurements BoF.
C. The Arbor ATLAS Initiative - Darren Anstee
Darren’s presentation is available at:
Richard Barnes (WG Chair) asked about IPv6 support.
Darren said most features of the product are v6 capable now, but there isn't much data, either because providers are not using the product for v6 or there are related bugs. Honeypot sensors are v4 only.
D. Home Network Monitoring - Renata Teixeira
Renata’s presentation is available at:
Robert Kisteleki, RIPE NCC, said he was going to run it at home and compare the results to RIPE Atlas.
Lorenz Colitti, Google, asked if the tool could detect broken IPv6 connectivity at the home, to help prepare people for IPv6 day.
Renata said the required measurements are taken, but no analysis is yet done.
Ingvar Mattsson, Google, asked if the slowness seen by Genouille might be related to the buffer bloat problem.
Renata replied that the netalyzer tool included in the profiler collects the relevant performance data and they also model the gateways with UPnP, so hopefully the analysis can be run later.
Daniel Karrenberg, RIPE NCC, gave some encouragement and asked for periodic status reports on the MAT-WG mailing list. He also mentioned how the results are relevant to Atlas measurements and a possible future Atlas-like switch could sit between the router and the modem to get less biased results.
E. NLNOG RING - Job Snijders, InTouch NV
Job’s presentation is available at:
Robert Kisteleki, RIPE NCC, commented that he would not want researchers to have shell access on his network.
Job clarified that only operators get access to your machine, which should be placed in the same LAN as other hosted machines.
Robert mentioned that it looks like Planet Lab and NL Lab.
Richard Barnes, WG co-chair, commented that RING may be more diverse then those networks.
Mohsen Souissi, AFNIC, asked if members are required to supply one machine per POP.
Job said one machine is sufficient, but more are welcome.
Richard Barnes said a list of current members would be appreciated.
Daniel Karrenberg, speaking as a “netizen”, said he would be interested in joining RING, but can’t allow portmap scans on his boxes.
From chat, Christiaan Ottow, Pine Digital Security, also asked about limiting the allowed commands.
Job answered that all the participants are from the land of the free, and restricting commands makes debugging harder.
Randy Bush, Internet Initiative Japan, asked for a clarification about the requirement for a server to be in the DFZ with its own ASN.
Job replied that the important thing is that the network and routing is controlled by the member, not another hosting provider.
F. Emile Aben, RIPE IPv6 Measurements
Emile’s presentation is available at:
Mohsen Souissi, AFNIC, asked if the NCC is interested in providers with permanent IPv6-capable networks or just those switching it on for D-Day?
Emile said they are interested in both.
Lorenzo Colitti, Google, asked about getting fav icons for the status page.
Emile said they will, but they are too small to notice MTU issues.
Lorenzo said they expect to see packet loss for the 6to4 return path instead of latency changes.
Lorenzo suggested trying to emulate the behavior of desktop OSes in addition to measuring them separately. For example, if you attempt first v6, then v4, you may experience a latency spike.
Geoff Huston, APNIC, commented that the tools are concentrated on the end client, but content providers would like to know if their stuff would work before they turn it on. So at APNIC they twisted the tool around to allow content providers to do it.
G. Christian Teuschel, RIPEstat - The Information Toolbox
Christian’s presentation is available at:
Richard Barnes asked that questions for RIPEstat be saved for the RIPE NCC Measurements BoF, thanks attendees, and closed the session.