Mastodawn

Soon to start, #RIPEduca. Expect some live-tooting about RIPE Atlas probes, Internet measurements and etc, if the proprietary bloatware Adobe Connect works.

Reminder: RIPE Atlas probes are a distributed network of probes that *you* can use to test Internet connectivity from many (10 000) vantage points https://atlas.ripe.net/

A must for any serious Internet professional.

RIPE Atlas - RIPE Network Coordination Centre

RIPE Atlas is the RIPE NCC's main Internet data measurement system.

These probes can do several sort of measurements, requested by users (you and me) : ICMP echo (ping), traceroute (UDP and ICMP), DNS, etc. Here, the current measurements: https://atlas.ripe.net/measurements/ #RIPEduca

RIPE Atlas is therefore a nice botnet.

Measurements - RIPE Atlas — RIPE Network Coordination Centre

RIPE Atlas, RIPE, Atlas

RIPE Atlas measurements can be submitted through several means: Web interface, and a nice API (several tools exist to use it). #RIPEduca

Kaveh Ranjbar now reminds us that RIPE Atlas probes have some limitations, to protect the network (no IP address spoofing, no arbitrary packets) and the people who host them (no arbitrary HTTP requests, because it is often monitored by the State or the Corporation). #RIPEduca

There are also special RIPE Atlas probes, the Anchors: larger machines, which can do the same things as the probes *plus* be the target of measurements (great for pinging the Internet).

#AFNIC one was among the first anchors installed.

ScriptFanix Oct 5, 2017

@bortzmeyer et si tu héberge une sonde Atlas, le ripe fait la supervision de ta connexion 😉
Avec même un rapport mensuel.

Adobe Connect does not work on Ubuntu and is quite unreliable on Windows. Those modern multimedia tools... #RIPEduca #wasBetterBefore

Rysiekúr (old account)Oct 5, 2017

@bortzmeyer Adobe Disconnect.

Now, RIPE Atlas probes and Internet Exchange Points, by Willem Toorop and Emile Aben.

"We will start with the first presentation" #captainObvious

A study of Internet connectivity in Canada

The Canada interconnectivity survey was done by analysis *existing* RIPE Atlas measurements (by default, measurement results are public).

Among the questions: how much of canadian traffic stays inside Canada? Between unicast machines (gov. Web servers for instabce) hosted in Canada? Between a probe in Canada and a anycast DNS server which has instances in Canada?

Usual problems of Internet measurements: traceroute with non-responding routers, routers replying with RFC 1918 (private) IP addresses, geolocation inperfect...

Ping @angristan about the ability of the US to shut down the Internet: some Canada traffic goes through the US.

#Internet #resilience #sovereignty

Result: 64 % of "canadian" traceroutes go through the US. (And these traceroute are often faster, which may be the reason: international links are upgraded more often; another reason is the lack of peering in the country)

Only 29 % of canadian governement sites are hosted in Canada. (And traceroute to them often go through the US.)

@angristan #RIPEduca

Half of the traceroutes to a #DNS #anycast root server which has an instance in Canada goes to Canada. New York or Chicago attracts most of the rest. Even worse for .ca name servers.

@angristan #RIPEduca

https://www.ripe.net/ixp-country-jedi

Emile Aben presents "IXP Jedi", a measurement of the actual use of Internet Exchange Points (like the France-IX) through RIPE Atlas measurements. Do they really "keep local traffic local"? [Insert picture of a very smiling user]

#RIPEduca #SchengenRouting #Snowden

A limit of all RIPE Atlas measurements, spotted by Emile Aben: probes vs. Internet users is heavily biased. Not enough probes in Africa, for instance (and "too many" in Europe). #RIPEduca

John Shaft (ジョン・シャフト) ✅Oct 5, 2017

@bortzmeyer In Asia too : only 15 probes in mainland China (and 28 in Hong Kong), 1 in Mongolia, 5 or 6 in Cambodia, less than 10 in Vietnam...

#RIPeduca #SchengenRouting

Example of #Finland: the most common path between two finlandeses probes is inside the country but not crossing an IXP. A minority goes through an IXP.

Example of #France: some FR-FR paths go through the US (and many go to another of the Five Eyes, such as GB) [Remember that Orange does not peer at a public Exchange Point in France]

Emile reminds us that #BGP does *not* take into account latency (unlike what you often read in media). The most important (often the only one) criteria for operators is cost, not latency or locality. #RIPEduca

Soon, the #DNS session at #RIPEduca. Expect no toots, I'm the one doing the talking. https://www.ripe.net/support/training/ripe-ncc-educa/agenda

RIPE NCC Educa::RIPE Atlas Agenda

You can participate for the full day or just log in for the sessions that interest you the most.

Resuming #RIPEduca https://www.ripe.net/support/training/ripe-ncc-educa/agenda with the tools to use RIPE Atlas probes. #Internet #measurements

RIPE NCC Educa::RIPE Atlas Agenda

You can participate for the full day or just log in for the sessions that interest you the most.

Christopher Amin on RIPE Atlas API https://atlas.ripe.net/docs/api/v2/

#RIPEduca #JSON #REST

#JSON #API #RIPEduca #IpreferXML

"Everything that you post is JSON. Everything that you get is JSON."

"Works in any language"

curl -q https://atlas.ripe.net/api/v2/measurements (see the measurements) curl -q https://atlas.ripe.net/api/v2/probes (see the probes) #JSON #RIPEduca

You can also create a new RIPE Atlas measurement (this time, you need an API key and credits) https://atlas.ripe.net/docs/api/v2/manual/measurements/types/index.html It even works with curl -X POST #RIPEduca

Create Measurements · RIPE Atlas API Manual v2

Did you know? The RIPE-NCC official tools to work with RIPE Atlas are named, Cousteau, Sagan and Magellan. #RIPEduca

By the way, @Keltounet where is the #GOlang implementation? https://github.com/RIPE-Atlas-Community/ripe-atlas-community-contrib #RIPEduca

RIPE-Atlas-Community/ripe-atlas-community-contrib

ripe-atlas-community-contrib - Repository for links towards tools written during hackathons, and a collection of contributions by the community of the RIPE Atlas visualizations, tools for analysing...

Ollivier Robert 🇺🇦😷🌈Oct 5, 2017

@bortzmeyer I still need to finish it :(

ZakMcKraken 🕵🥃Oct 5, 2017

@bortzmeyer you prefer #xml ? ibm has a solution for you https://www.ibm.com/support/knowledgecenter/SS9H2Y_6.0.1/com.ibm.dp.xb.doc/json_jsonxconversionexample.html 🙌😅 #JsonX

IBM Knowledge Center

Massimo Candela about the other RIPE Atlas API (the streaming one) and how to use it.

Demo of LatencyMON at #RIPEduca Here, my blog https://atlas.ripe.net/measurements/2060427/?lm2060427.start=1507156110&lm2060427.stop=1507204110&lm2060427.measurements=2060427&lm2060427.groups=2060427-10892,2060427-11334,2060427-11213,2060427-11421,2060427-2195&lm2060427.groups-component=2060427M10892L2060427M11334L2060427M11213L2060427M11421L2060427M2195&lm2060427.groups-type=single-probe,single-probe,single-probe,single-probe,single-probe#!latencymon

And now TracerouteMON https://atlas.ripe.net/measurements/9412526/#!tracemon (here to North Korea) #RIPEduca

Don't hesistate to click everywhere on the TraceMON picture : lots of features! #RIPEduca

TODO: play with the RIPE Atlas streaming interface one day. https://atlas.ripe.net/docs/result-streaming/ Seems very cool. #RIPEduca

RIPE Atlas Result Streams - RIPE Atlas — RIPE Network Coordination Centre

RIPE Atlas, RIPE, Atlas

Now, measuring service quality with RIPE Atlas at #RIPEduca. With Christian Kaufmann and Vesna Manojlovic.

Ping #ARCEP

Christian Kaufmann (also president of the #FranceIX board) talking (prerecorded video, no interaction) at #RIPEduca.

Measuring the difference between performances with #IPv4 and #IPv6.

"I used the API because I didn't want to click 500 times on a GUI" #RIPEduca #CLI

Shorter AS count with IPv6. Not much difference in RTT with IPv4. #RIPEduca

Sylvain Lesage Oct 5, 2017

@bortzmeyer what does it mean ?

Stéphane Bortzmeyer Oct 6, 2017

@severo_bo Probably this is because everyone peers with Hurricane Electric, so everyone is close (AS-wise). #IPv6

Vesna Manojlovic now speaking for webmasters: "measuring reachability of your Web server" with RIPE Atlas #RIPEduca

Atlas was more focused on layer 3. But users requested L7 measurements.

Ping @Natouille, @aeris and the others at #ParisWeb

Limitations of HTTP measurements: only towards RIPE Anchors, partly for ethical reasons (forbidden Web sites) partly to avoid LOIC-type dDoS.

The trick: "TCP ping" with the traceroute measurements towards port 80, packet size zero.

Just implemented "TCP ping" in my atlas-traceroute tool https://github.com/RIPE-Atlas-Community/ripe-atlas-community-contrib/commit/f5c4c181273f338e48705fbf7731a05b22b4e344 https://atlas.ripe.net/measurements/9716802/ https://labs.ripe.net/Members/stephane_bortzmeyer/using-ripe-atlas-to-debug-network-connectivity-problems #RIPEduca

New options (TCP, size, port), to do "TCP Ping" · RIPE-Atlas-Community/ripe-atlas-community-contrib@f5c4c18

Now, session on security at #RIPEduca. Laura Roberts speaking on #Tor and its use of the #DNS. (Note there was two women speaking during this day. Not too bad.)

Executive summary: Tor exit nodes leak info through their #DNS requests. 1/3 of them use Google Public DNS! #RIPEduca

RIPE Atlas probes were used to infer AS paths, through traceroutes. There are probes in half of the AS where there is a #Tor node. Same thing for just the exit nodes. #RIPEduca

Simulation: the best solution is (obviously) a #DNS resolver on the #Tor exit node. #RIPEduca

John Shaft (ジョン・シャフト) ✅Oct 5, 2017

@bortzmeyer Il faudrait placarder le RFC 7626 chez tous ces admins (irresponsable) de nœuds de sortie et leur faire apprendre par coeur 😥