RIPE 64

RIPE MAT Working Group Minutes

RIPE 64, Ljubljana
Thursday 19 April 2012, 16:00 – 17:30

Status: Final

Co-Chairs: Richard Barnes, Christian Kaufmann, Ian Meikle
Scribe: Rumy Spratley-Kanis, RIPE NCC
Chat Monitor: Alex Band, RIPE NCC

A. Introduction

The MAT Working Group Co-Chair Ian Meikle opened the session and welcomed the attendees on behalf of all the chairs. He thanked the scribe, chat monitor and stenographers and asked the participants to state their name and affiliation in case of questions/comments at the microphone. He also mentioned that there was a slight change to the agenda and that Daniel Karrenberg's talk was moved to the end.

B. Mobile Broadband Measurement - Amund Kvalbein, Simula

The presentation is available at:
https://ripe64.ripe.net/presentations/172-Mobile_Broadband_Measurements.pdf

Martin Levy, Hurricane Electric asked if there was a way to measure whether it's the local 3G radio signal, the back haul, or the infrastructure the back haul goes into. He added that there are multiple components and wondered if it was even possible to guess.

Amund Kvalbein confirmed that it is difficult, but guesses were possible by working with operators. He said that you can look at the causes of drops and delays in queues with limited certainty without getting data from the operators.

Martin asked if the data was public, or made public when they did it.

Amund replied that they are hoping to make it public very soon, it is one of their goals. He added that there are operators involved in this who will have to agree.

Geoff Houston, APNIC, said that what they (Amund et all) are trying to measure is the quality of the product and that it is not their problem to fix mistakes. He suggested that maybe they should be using the probes more than as ping targets, but as delivering a service. He added that, when measuring v6, it is interesting to note that there are two parts to a connection: the client and the other end. If you become the other end, the client doesn't need to measure the quality. The server can, because it's the same connection. He added that you can get a lot of data if you get the clients to send TCP sessions every so often and the server will give you RTT and the stability of RTT and jitter. What they're really after is how well do data applications perform. And that is TCP and how well does TCP perform? Geoff suggested that if they could get the nodes and servers and start instrumenting the server they actually got something useful for their regulators and customers and all the rest.

Amund replied that this is a very good point and that he was perhaps not clear about this, but that the servers that they do the pings to are their servers, they have both ends and want to do other things than just pinging.

Geoff Huston said that that was the point.

Amund said that they want to do TCP.

Geoff replied that the TCP state gives a lot more information than just a ping.

Amund thanked Geoff for this good input and said they would try to do this. Part of the reason they started with just ping for the first set of measurements is because the subscriptions they had were limited. They could only send limited amounts of data.

C. Useful and Cool New Things in RIPEstat and RIPE Atlas - Vesna Manojlovic, RIPE NCC

The presentation is available at:
https://ripe64.ripe.net/presentations/221-atlas-ripestat-mat-v8.pdf

There were no questions.

D. APNIC Measurement of IPv6 Using Paid Advertising - George Michaelson, APNIC

The presentation is available at:
https://ripe64.ripe.net/presentations/227-RIPE64-mat.pdf

The chair thanked George Michaelson for sticking to the charter of the working group.

Randy Bush asked about privacy.

George replied that they respect privacy and do not have any knowledge of where they went.

Randy Bush said that they are collecting, and no better than Google.

George replied that the advert says "thank you for helping us measure v6".

Randy commented that they are collecting more than that.

George replied that they don't know the identity.

Randy said that his home IP address doesn't change.

George replied that it is a fair question, but not a real question. It has to be considered. He also noted that the class of information is worth the risk and they make a strong statement.

Randy said that the end justifies the means is not a good argument.

George replied that he feels they are within bounds.

E. MLab Update - Tiziana Refice and Meredith Wittaker, Google

The presentation is available at:
https://ripe64.ripe.net/presentations/220-RIPEMeetingMATWG-04192012.pdf

There were no questions.

F. RIPE NCC Measurement Strategy 2012/2013 - Daniel Karrenberg, RIPE NCC

The presentation is available at:
https://ripe64.ripe.net/presentations/150-20120419-ncc-mat-wg-dfk.pdf

Dave Wilson, HEAnet, said that as a TTM host, while he appreciates the service has to stop, he has one regret, or request. He was wondering if there is any way to salvage some of the great work that has been done, like the hardware already in place and hosted or the software or historical measurements. Maybe there is some group out there interested in taking some of that on. He added that he would be grateful if the presenter would be interested in this as well.

Daniel Karrenberg replied that they are interested in preserving the historical data and making it available. He added that there is a data store at the NCC and the intention is to put all the data there. He continued to say that the current thinking is to make the software available as is under some sort of open source licensing. He added that as far as hardware is concerned: they are not going to collect the hardware back. The remaining life in most of the hardware out there is not great and it's impossible to get replacements, upgrades or spare parts, which was one of the reasons that pushed towards canceling TTM.

Dave replied that that is good news and he appreciates it. He added that it would be great if there were a single home, some group out there that is doing something similar and might want to take it on.

Daniel Karrenberg suggested that people contact him, or Dave, or get on to the mailing list.

Gert Döring, SpaceNet, mentioned that, having two boxes, what he is worried about is that they base their SLA measurements that they guarantee their customers on the TTM network because that's a neutral entity. He added that he understands that this should be possible with the anchor boxes, and he hopes that they can get one or two anchor boxes before shutting down the TTM
boxes so that the transition will go smoother.

Daniel Karrenberg replied that that is the intention. Or to keep running the Atlas things. He added that he understands that there are a few TTM hosts with the same issue, Gert is not the only one. He continued to say that he pledges to help them as well as they can to have continuous SLA monitoring if the anchor boxes are suitable or to take over the thing locally if the anchor boxes are not suitable. He finished by saying that he would be very happy if people offer their help in developing the anchor boxes.

Gilles Massen, RESTENA, via chat, mentioned a question on the MAT-WG mailing list concerning DNSMON and said it was the only way to monitor the quality and reachability of their DNS service, specifically as many of the nameservers are anycast clouds. He said that what mattered was having easily interpreted graphs and overviews, with the ability to drill down on servers or probes, historical information (within the graphs) and detailed (raw) data.

Daniel Karrenberg replied that he had read the mail and that all that is something they want to preserve.

Peter Koch, DENIC, mentioned that they are a paying DNSMON customer and while they know about a firm end date for the TTM box there doesn't seem to a firm start date for an Atlas-based DNSMON service, and that this is a concern for him.

Daniel Karrenberg replied that there is no intention to stop the DNSMON service, only the TTM service, and that he could draw his conclusion from that.

Peter Koch said that that was not a satisfactory answer.

Daniel Karrenberg replied that they intend to continue the DNSMON service, maybe change it a little, but there will be continuance. That is the position right now.

Peter Koch said that he understands but he would rather have a firm commitment rather than an intention.

Daniel Karrenberg replied that there is a commitment of a contractual relationship and it will be honoured.

Ian Meikle, MAT-WG co-Chair, mentioned the mailing list again and encouraged the discussion to be continued on the list. He encouraged Daniel to respond to the post about DNSMON on the list.

Daniel replied that he responds to all mails on the list and looks forward to seeing more.

Ian encouraged him to address some of the issues, or make a summary of this talk so that they could start a discussion on that.

Z. AOB

There was no other business to attend to.

The Chair thanked everyone for attending and closed the session.