Connect Working Group Minutes RIPE 73

Wednesday, 26 October 12:00 - 13:30
WG Co-Chairs: Remco van Mook, Florence Lavroff
Scribe: Emile Aben

1. Welcome

Remco opened the session.

Remco asked if people avoided the Address Policy Working Group and about half the room raised their hand.

Remco asked if people wanted to see interconnect presentations and about 3/4 of the room raised their hand.

2. Hackathon

Pinder - Matthew Walster, Fastly

The presentation is available here:
https://ripe73.ripe.net/presentations/67-Pinder.pdf

There were no questions.

Remote Peering Jedi - Vasileios Giotsas, CAIDA

The presentation is available here:
https://ripe73.ripe.net/presentations/113-The-remote-peering-jedi.pdf

There were no questions.

3. Illegitimate Source IPs At IXPs

Franziska Lichtblau, TU Berlin

The presentation is available here:
https://ripe73.ripe.net/presentations/12-Illegitimate_ips_at_IXPs_ripe73_franziska_lichtblau.pdf

Andrei Robachevski (ISOC) asked if they compared their results to other spoofer projects.

Franziska answered they didn't do a full comparison of their results with other spoofer projects, they just used them for verification.

Andrei asked how they tackled multihomed customers that emit traffic through a provider, but don't announce the address space through it.

Franziska answered they did some manual filtering, and are limited by what they can see publicly.

4. Exporting Flow Telemetry for Fun and Profit

Nick Hilliard, INEX

The presentation is available here: https://ripe73.ripe.net/presentations/138-inex-ripe-connectwg-2016-10-26.key(PDF not uploaded)

Eric Nghia Nguyen-Duy, AMS-IX, asked about the scalability of the setup Nick described.

Nick answered they did it with a full INEX sflow feed and a mapping list of 1,000 entries which consumed 8-10% of a virtual CPU. Nick thought it should scale reasonably well and handle large IXPs, also because reducxing can be done at another machine.

Filippe Duke, Netassist, asked how accurate the solution was.

Nick answered he thinks it's 100% accurate if the mapping between MAC address and port is accurate.

Steve Nash, Arbor Networks, offered to collaborate with IXP participants that are also Arbor collectors.

Nick appreciated that.

Paolo Lucente, pmacct, replied to Eric by saying that scalability is not a problem. He offered help to AMS-IX on that.

5. Scaling BIRD Route Servers

Benedikt Rudolph, DEC-IX

The presentation is available here:
https://ripe73.ripe.net/presentations/115-e-bru-20161026-RIPE73-scaling-bird-routeservers-final.pdf

Peter Hessler, OpenBGPd, asked about a spike and a trend in the comparison of executing times.

Benedikt explained that for a certain exact amount of peers, it seemed the process took a break and only resumed after 50 seconds. He said that with more repetitions that signal might go away, so maybe it's an outlier.

Martin Levy, Cloudflare, asked how long Benedikt spent on this research.

Benedikt responded this was 2-3 weeks of continuous effort.

Martin Levy wondered if he could have better worked on multi-threading in BIRD instead.

Benedikt answered there were other benefits from working on this.

Ondrej Filip, CZ.NIC, asked which version of BIRD was used.

Benedikt answered that BIRD 1.6 was used.

6. Preview: Show Me the Money (The Unique Financial Challenges of Smaller Internet Exchanges)

Leslie Carr, Clover Health/SFMIX

The presentation is available here:
https://ripe73.ripe.net/presentations/120-Show-Me-the-Money.pdf

There were no questions.

7. Feedback About the Session and Closure

Remco asked for feedback about the session and closed the session.