RIPE 67 MAT Working Group Minutes

Measurement, Analysis and Tools (MAT) Working Group Minutes

Date: 17 October 16:00 – 17:30
Chair: Christian Kaufmann
Scribe: Suzanne Taylor Muzzin

A. Introduction

Christian Kaufmann welcomed the audience and thanked the scribe, stenographers and chat monitor. He asked whether the minutes from RIPE 66 could be approved; attendees approved the minutes with no proposed revisions.

B. Probing the Internet through Smartphone-based Crowdsourcing
Enrico Gregori, IIT-CNR

This presentation is available at:
https://ripe67.ripe.net/presentations/196-Portolan.pdf

Alex Shcherbakov, Garant-Park-Telecom, asked for clarification about whether they allow people to look into these statistics, like with RIPE Atlas.

Enrico responded that yes, of course they do allow that and that they are a research institution, so the data is completely public and the methodology is available.

Alex asked whether, if he manages his mobile network and installs the app, he could run measurements of his own.

Enrico replied that yes, that is what’s defined as a user measurement.

Benno Overeinder, NLnet Labs, asked whether they provide a similar interface to the one used by RIPE Atlas.

Enrico responded that they are still working on the interface. He added that it’s not a problem to find a UI that’s accepted by the community, and that’s the reason for crowdsourcing.

Daniel Karrenberg, RIPE NCC, joked that he found it interesting that RIPE Atlas has already received “legacy status”. He asked Enrico what his experience has been so far with people who download the app and do measurements for week or two; specifically, he wondered how many of those drop out.

Enrico replied that although he doesn’t have exact statistics, his feeling is that they’re still at partial numbers. But he said that if a measurement is useful for users, they tend to leave the app running, because it uses less than 1% battery power and so that doesn’t affect most people’s performance. He said they will have to wait at least a year to see where it goes, but they will continue to develop the services that people find interesting.

Daniel suggested that one idea would be to advertise the app among the RIPE Atlas probe hosts and then do some calibration with them at home. He added that if they have a probe at home and have the app running on their phone over WiFi, they could see whether there are differences in the performance based on the WiFi.

Enrico responded that it sounds like an interesting idea and he will follow up with Daniel.

Tiziana Refice, Google, asked whether the code is available and open source, and whether the raw data is available, because she couldn’t find it on the website.

Enrico responded that, so far, the app is free but not open source, and that the raw data is available. He added that they’re open to discussion and are interested in what the community wants.

Tiziana asked about the measurement infrastructure being used to run the measurements against the servers.

Enrico responded that it’s what they call regional Italian infrastructure, which is the Internet that you cross when start from an Italian node and your target is located in Italy.

Tiziana asked if that meant that basically, their servers are in their network.

Enrico confirmed that yes, so far they’ve only used this one, but added that they could do much more.

Christian asked that all other questions be taken offline in the interest of keeping to the schedule, and thanked the speaker.

C. RIPEstat Update

Vesna Manojlovic, RIPE NCC

The presentation is available at:
https://ripe67.ripe.net/presentations/351-ripe67-mat-wg-ripestat-update-v1.5.pdf

Florian Streibelt, Technical University Berlin, commented that Vesna should have mentioned RIPEstat’s excellent mobile interface.

Vesna thanked him and responded that they prefer users to praise them rather than praising themselves. The audience applauded.

Tiziana Refice, Google, said the she is really happy with the integration of M-Labs data into RIPEstat. She said that she works on M-Labs and since the integration, many more people are looking at the M-Lab data via RIPEstat. She also asked whether the RIPEstat team is looking into things like possibly adding network events to target events on a timeline, which would be useful.

Vesna referred to Emile Aben, RIPE NCC, the “events specialist”, and asked if there was anything he wanted to say.

Emile responded that he would love to add network events, too.

Tiziana asked if that is in the roadmap, and Vesna responded that she will
make sure it appears and contact Tiziana later to find out exactly how she’d like to see it displayed.

Daniel Karrenberg, RIPE NCC, wondered if he understood Tiziana correctly, and asked if the suggestion is to line up widgets on a timeline so you can see the progression of an event. He added that there was an idea several years ago to do that and to add comments to it in order to document something that happened, and basically create a story within RIPEstat. He asked who in the room would use something like that, and not many people put up their hands.

There were no further questions.

D. Infrastructure Geolocation Using RIPE NCC Data Sets

Emile Aben, RIPE NCC

The presentation is available at:
https://ripe67.ripe.net/presentations/341-2013-10-ripe67-router-geoloc-emile-aben.pdf

Emile asked who in the audience liked the idea presented in his talk of providing more data and tools around geolocation. A large number of people in the audience raised their hands to show support.

Jared Maunch, NTT Communications, asked whether Emile has a list of the ones for which he already knows their locations. He said that they have most of them documented on their website for the route community, and asked if there is a central repository or place where Emile will post them, along with the ones he doesn’t know, so the community could provide feedback.

Emile said he wasn’t sure what the best interface was, but that it’s something he’s interested in and find out how to make it as easy as possible for people.

Mike Hughes, Ethernorth Consultancy, asked whether Emile is concerned that people will try some interesting policy routine so they can spell things and draw things on the map.

Emile responded that he’d thought about holding a contest for the most beautiful visual traceroute out of this.

Mike said that, in terms of serious research, he’d been doing some work and it would be really useful for people at smaller regional exchanges to see things such as hairpinning.

Wilfried Wöber, UniVie/ACOnet, commented that he has a RIPE Atlas probe and loves the idea, and would like to see something like traceroute play that’s as fancy as BGPlay. But he said the question he always has, in the context of the RIPE Database and starting to do something with geoloc tools, etc. is about the fact that the people who have the authoritative, credible information are often not the ones who have access to reverse DNS or to update their address block in the hierarchy of their ISP. He asked who enables people in the know to put the data in the correct place to be accessible.

Emile responded that, for now, the idea is to not constrain access to things like reverse DNS. He suggested you could use triangulation, for instance, to see where stuff is approximately, to within a few hundred kilometres. He also said that anybody in an organisation could put information for IP addresses in.

Wilfried rebutted that, in that case, it is not put into the authoritative place, but a web interface that would be provided by the RIPE NCC. He added that he doesn’t understand how Emile thinks he could collect data from people and still be aligned with policies in a company or exchange point, etc.

Emile admitted that someone could try to provide information for someone else’s network.

Daniel Karrenberg, RIPE NCC, stated that one of the things they’re thinking about is that the way to start this is to be open, where anyone can contribute to this. But he suggested that, maybe in the UI, you could rate people on the quality of their data, and then weight the data. He added that they’re already working on traceroute play-type tool but it is not ready for prime time yet.

Christian asked everyone to take these sorts of discussions to the mailing list. The audience applauded that idea.

Kay Rechthien, ECIX, suggested that the whole geolocation thing is rather broken right now, and to take it to a higher level (for example, this year a network is in Hamburg and next year somewhere else), would be useful.

Emile responded that, in the Database Working Group session, Zoltan Szamonek, Google, gave a presentation in which he suggested a data feed that would solve the edge case.

Robert Kisteleki, RIPE NCC, added that they just wanted to pitch this idea to the community and get a feel for whether it’s a good idea or not. He added that it’s clear that the majority of people do think it’s a good idea and that everyone could expect lots of follow-up discussion and questions on the mailing list and RIPE Labs, etc.

There were no further questions.

E. Mapping Dutch Critical Infrastructure

Razvan Oprea, RIPE NCC

The presentation is available at:

https://ripe67.ripe.net/presentations/355-Presentation_Mapping_the_Dutch_Critical_Infrastructure.pdf

Andrei Robachevsky, ISOC, asked to what extent Razvan mapped the kind of external facing interface vs. corporate infrastructure (such as corporate databases, email system, IT systems, etc.), which Razvan likely wouldn’t have discovered using the method he did.

Razvan responded that the only interfaces he looked at were the AAAA and MX records. He said there are very few cases in which you could avoid having your email system go through the MX records. He added that he understands the data is not complete, but that whatever data they used, they checked one by one. He also said it could be extended further if you had privileged access, but he didn’t want to go down that route for this project.

There were no further questions.

F. RIPE Atlas Update

Vesna Manojlovic, RIPE NCC

This presentation is available at:
https://ripe67.ripe.net/presentations/350-ripe67-mat-wg-atlas-update-v1.5.pdf

Christian asked Vesna to put the questions she has from people not in the room on the mailing list. Vesna replied that she would do that.

Stéphane Bortzmeyer, AFNIC, said that, regarding big questions about the future of RIPE Atlas, there’s the question of using RIPE Atlas for commercial use, such as selling results collected by RIPE Atlas, and there’s been discussion about whether we need to restrict the way in which the data is used and be more explicit about that. He advised people to go back to the mailing list if they have something to offer on that topic.

Vesna thanked him and said she forgot to mention that, but that she would add that to the questions circulated on the mailing list.

Kaveh Ranjar, RIPE NCC, said he wanted to clarify the situation about the terms and conditions, because RIPE Atlas didn’t have clear terms and conditions until now, and that the RIPE NCC will be sending some suggestions to the mailing list to get feedback from the community about what they want. He said they are waiting to finalise these but will be sending them to the list in the near future.

Robert Kisteleki, RIPE NCC, wanted to thank Stéphane for the workshop he led on RIPE Atlas on the Monday before, which the RIPE Atlas team had nothing to do with. The attendees applauded.

Vesna commented that she brought a RIPE Atlas anchor to the meeting to show the community. She added that there are also plans to install them at all K-root locations and RIS route collectors. She said that, since the RIPE NCC has a K-root location in Athens, she wanted to ceremoniously present an anchor to the meeting host, Andreas Polyrakis, GRNET.

Andreas thanked Vesna and thanked the attendees for another great meeting. The audience applauded.

Kaveh Ranjbar, RIPE NCC, also thanked Andreas and GRNET for being an excellent host. There was more applause.

Z. AOB

Christian also thanked Vesna and GRNET.

Mikael Abrahamsson, T-Systems, said he pitched this idea at the IETF in Stockholm four years ago and that no one took it up, so he would pitch it here again. He said there is a lot of information stuck in his TCP stack that he can’t get to, such as every packet loss, round-trip times of all connections, and all of this over time. He said if he could somehow statistically represent this, it would tell him a lot about his what’s going on with the network and what the user experience is. He suggested this is something you could have on all hosts and that it could be non-intrusive. He said he just subscribed to the mailing list and was just pitching this idea.

Christian thanked him for his comment.

Randy Bush, Internet Initiative Japan, said that in the 1990s there was an app called Vital Signs that ran on computers and monitored the stack and did passive monitoring on TCP connections, and was designed for the user – so this has already been done.

Hans Petter Holen, Visma, said that they are running a commercial product that does most of what Mikael is talking about, called EdgeSight by Citrix.

Tiziana Refice, Google, mentioned another program that will allow this on Linux machines, and said she could talk to Mikael about it more later.

Christian apologised for having to cut off so many people due to a lack of time, and asked everyone to take these discussions to the MAT Working Group Mailing List. He thanked everyone for attending and closed the session.