RIPE 69 MAT Working Group Minutes

Date: 5 November 2014, 14:00-15:30
Chair: Christian Kaufmann
Scribe: Suzanne Taylor

A. Introduction

• Welcome
• Scribe
• Jabber
• Agenda

Christian welcomed everyone and asked whether the minutes from RIPE 68 could be approved. There were no objections and the minutes were approved.

Christian said that MAT Working Group Co-chair Richard Barnes was unable to attend RIPE 69, but that they've been working on bullet points about how to appoint MAT Working Group Chairs. He urged everyone to look at the MAT Working Group Mailing List to review the proposal and to come forward if they want to be considered as Chair to help with RIPE 70 in Amsterdam.

B. On the Suitability of Two Large-Scale Internet Measurement Platforms

Randy Bush, Internet Initiative Japan

This presentation is available at:
https://ripe69.ripe.net/presentations/92-141105.ripe-atlas-nlnog-compare.pdf

Randy compared two Internet measurement platforms: RIPE Atlas and NLNOG RING nodes. He described how he pinged both RIPE Atlas probes and NLNOG RING nodes and saw events that affected each of them differently. He also looked at global average RTTs to both platforms, and found an average of 0.5ms difference between them. He also showed how RIPE Atlas probes and NLNOG RING nodes result in no more packet loss than dedicated BSD servers.

Philip Homburg, RIPE NCC, said he was surprised that the graph of different flow IDs shows RTTs for RIPE Atlas probes being affected differently than those for the NLNOG RING nodes. Randy said that he purposely showed the interesting results.

Vesna Manojlovic, RIPE NCC, suggested Randy could repeat his measurements and distinguish between RIPE Atlas probes, which she said are not really meant to be used as measurement targets, and RIPE Atlas anchors, which are meant to be used as targets. Randy responded that the regular probes work well as targets, but agreed there are a number of different things they could look at in the future, such as different probe versions. He said that if someone is interested in helping him with this project, they should get in touch.

Daniel Karrenberg, RIPE NCC, asked Randy to please send his code to the MAT WG Mailing List. He also asked whether Randy looked at how many RIPE Atlas probes out of his random 250 sample were behind NATs. Randy said he believed someone did look at that. Daniel said that users can now find out whether probes are behind NATs thanks to the tagging feature, and that this would be useful to know, because there's a high probability that the NAT is the one answering and not the probe.

Jen Linkova, Google, asked whether Randy did these measurements over IPv4, and he confirmed that. She suggested they measure IPv6 as well. Randy responded that there aren't many targets available, but that it's a good point and NLNOG RING nodes are all supposed to be IPv4 and IPv6 capable. Vesna added that all RIPE Atlas anchors are also IPv4 and IPv6 capable.

Christian said that he finds it useful to be able to compare results between RIPE Atlas probes and other measurement sources, rather than just between the probes themselves, in order to have an objective view of reality. Randy responded that this is not exactly the case, but it is a sort of recalibration, so you can at least know whether everyone else is seeing the same thing, even if it's not the objective truth.

There were no further questions.

C. The Value of WLAN Measurements for the R&E Community

Brook Schofield, GÉANT Association

This presentation is available at:
https://ripe69.ripe.net/wp-content/uploads/presentations/91-eduroam-and-Atlas-RIPE69.pdf

Brook gave an overview of eduroam, a roaming service for educational institutions that is similar to other wireless hotspots but runs via an overlay network. He explained some of the different ways they've tried to measure the service using RIPE Atlas probes.

Ondřej Caletka, CESNET, said he liked the idea of using probes to measure eduroam, but that some probes are located at data centres and other insulated locations, and there are only a few access points. He said that for a proper measurement you would need a probe for every access point, which is not scalable.

Brook said that they want to scale this to every campus with eduroam, but that not all campuses run their own network. He said that they believe one probe per controller should be sufficient to verify the network's operation, and that whether an access point is visible from the data centre is something to be explored. He continued that they'd like to be able to turn on wireless LAN measurements to see whether they can spot eduroam access points.

Ondřej asked about the test accounts dedicated for these measurements and said that the test accounts present limitations and some real accounts would be required to access the networks.

Brook said that they debated this and then discussed with RIPE NCC developers whether to enable passwords or not. They decided that there was some value in doing full connectivity, looking at whether the network supports IPv4 and IPv6 and whether the IPv6 networks are behind NATs, etc., and whether they could test for the service's functionality. He said the downside to test accounts is people using their own personal accounts, but that it's difficult to get around this and that they think there's still some value in deploying passwords.

Dave Wilson, HEAnet, said he really supports this work and that this is something that many communities can benefit from, and this is exactly the kind of thing for which RIPE Atlas should be used.

Daniel Karrenberg, RIPE NCC, said that the RIPE Atlas team has looked at whether the v3 RIPE Atlas probes can be used to do this, and that it's a matter of priorities, and these should be discussed in the MAT Working Group. He said he believes the RIPE NCC can add value to this project and encouraged everyone to give their feedback about what features and capabilities the community wants, and how to prioritise these. He asked everyone to look at the RIPE Atlas Roadmap and to give their feedback to the RIPE NCC about what is important to them. He explained that the whole idea behind RIPE Atlas was to provide a common measurement platform that the whole community can use, which in turn makes the network stronger.

Kaveh Ranjbar, RIPE NCC, said that the RIPE NCC will create an implementation proposal for the community and the MAT Working Group Chairs within the next few weeks.

Bartosz Musznicki, INEA S.A., asked about whether an opt-in/opt-out option should exist for probe hosts who don't want their probes used for this purpose.

Brook responded that they would not scan any SS IDs or anything else on the network, just as the probes don't disclose any information about the probe host's network.

Daniel responded that the RIPE NCC will never enable wireless capabilities on the probes without the probe host's explicit permission.

Brook mentioned that the request for wireless measurements was already on the RIPE Atlas Roadmap before he asked for it, and encouraged everyone to communicate their own ideas about how to use this feature on the MAT Working Group Mailing List.

There were no further questions.

D. OpenIPMap (Crowdsourcing Internet Infrastructure Geolocation)

Emile Aben, RIPE NCC

This presentation is available at:
https://ripe69.ripe.net/wp-content/uploads/presentations/83-2014-11.emileaben.ripe69.openipmap.pdf

Emile gave an overview of OpenIPMap, an open crowdsourced project to map IPs and hostnames to geographical locations, which allows users to map traceroutes. He explained some of the challenges involved and asked the community for feedback as the project expands beyond it's original small group of volunteers and explained that OpenIPMap is now included in RIPE Atlas' new traceroute results interface.

Kaveh Ranjbar, RIPE NCC, asked for feedback about how useful the project is and how many resources the RIPE NCC should put into the project.

Alison Wheeler, The Creative Organisation, asked (via chat) whether it is possible to use the “LOC” type in a DNS record for a domain, which is user-created, and whether Emile has thought about getting users themselves to crowdsource their locations.

Emile responded that DNS LOC records are looked up and the graphical interface exposes this. He said that self-identification is more involved at the edges, which is no the scope of this project.

E. RIPE Atlas Data Visualisations

Vesna Manojlovic, RIPE NCC

This presentation is available at:
https://ripe69.ripe.net/wp-content/uploads/presentations/99-RIPE69DataVisualisation.pdf

Vesna showed some of the new data visualisations available using RIPE Atlas data, including some that showed BGP data, geolocation data, real-time DNS data, and many others. She asked the community to share their own favourite data visualisations and think about how to use RIPE Atlas data creatively.

Kaveh Ranjbar, RIPE NCC, expanded on the new DNS real-time data visualisation the RIPE NCC created and explained that there is a movie available that shows the AMS-IX nodes going down and coming back online when AMS-IX recently renumbered.

George Michaelson, APNIC, asked about getting the RIPE Atlas network to the extent that a visualisation of RIPE Atlas probes would outline the world's population.

Daniel Karrenberg, RIPE NCC, said that he has worked on such visualisations and will have them available at the next RIPE Meeting.

Thomas King, DE-CIX Management GmbH, thanked Vesna for all the great work and said he was very interested in the IXP tool, and asked if it could be open sourced.

Martin Levy, CloudFlare London, LLC, said that he is inspired when he sees interesting, visually stunning data visualisations and encourage the community to come up with their own visualisations based on RIPE Atlas before the next RIPE Meeting.

F. Python and RIPE Atlas

Vaibhav Bajpai, Jacobs University Bremen

This presentation is available at:
https://ripe69.ripe.net/presentations/121-README.txt

Vaibhav gave an overview of her work on measurements within Python, including examples of the actual code used to retrieve RIPE Atlas data to look at the distribution of RIPE Atlas probes by ASN index.

Christian said he found it interesting to see how much was possible with so few lines of code.

There were no further comments or questions.

Z. AOB

Christian reminded everyone to please comment on the proposal about replacing MAT Working Group Chairs and to think about nominating themselves.

He thanked everyone for attending and closed the session.