NTAC Peering Working Group

["Schmiedt, Dan" <>]

  Notes from 6/20 NTAC Peering & Routing SIG call

 

1. Agenda Bash

1. Wanted to query the group on Remote Triggered Black Hole Routing (RTBH) – Jeff Bartig

Update on peering and I2PX

1. I2PX – working with capacity augments. 
2. NANOG 88 last week in Seattle.  Met with Fastly, Zayo, Oracle, discussed capacity augments with all.  Fastly, Zayo  – replacing 10G with 100G. 
3. Oracle – 2 new regions in US – San Jose, Chicago, in addition to Ashburn and Phoenix.  They operate each of these as islands.  Their backbone is only used for their own internal traffic, no transit traffic on their backbone.
4. Many routing security discussions at NANOG.  There was a roundtable to discuss current state vs upcoming capabilities like ASPA.
5. AWS gave a presentation at NANOG on their network infrastructure.  Noted that they have transitioned to all custom hardware, even at the edge, which was historically standard vendor.  They have 1RU chassis 32x400G. 2RU for backbone, encryption on all backbone links. 

1. David Sinn used to work on this project and is no longer under NDA if anyone wants to discuss.  David noted that one reason for doing this was that this allows them to not be limited to CPU in a specific network box.

Lightning talks – Noted that ISC deprecated ISC DHCPd.  Warned that it will likely one day stop working after an OS upgrade.  ISC promoting updating to new Kea software. Network Weather Update – Trends and Coming Events

1. Summer so things are quiet.  Brief conversation at NANOG with Amazon Prime about Thursday night Football.  They have been working to be sure CDNs have appropriate capacity.  I2PX is more complicated because regionals also do CDN peering, so it’s hard for them to understand where to send content.  Jeff is brainstorming on how to help them understand how to best peer with the R&E world.

Internet2 Network Updates

1. Project to upgrade backbone routers to IOSXR 7.9.1.  3 outage windows scheduled this week to upgrade 1/3 of routers each day.  Problems this morning causing backbone interfaces being down, so have backed out.  Working with Cisco to understand what went wrong.
2. Installed Arista at MANLAN to handle 400G.  Plan to phase out QFX1002 and replace with the Arista for 400G support.  Similar plan for WIX as well as Boston, where a new 400G exchange will soon be added.
3. Cloudconnect: now have GovCloud with Azure and Oracle in Ashburn.  Oracle also in Phoenix/Dallas.
4. What was driver for IOSXR upgrade?  Looking for new features, like label depth, expanded FIB, TI-LFA fixes. 
5. Any weird issues with Licenses?  Jeff wasn’t sure but though they might have had issues.  Licensing on each chassis as opposed to license server.

Any other business?

1. RTBH routing: some lengthy threads on this topic.  When deployed NGI – push was to minimize new features.  R&E network does accept routes with RTBH community set.  Concern is that I2 does not currently do strong filtering from peers as compared to participants.  Possibility that an incorrect Black Hole route could be announced.  Jeff asked if the group shares concern and if I2 should consider not accepting RTBH routes.  Jeff’s gut feeling is that no one would notice and would reduce a risk.

1. No significant input from group pro or con.
2. Jeff: RTBH can be anything from host route up to prefix size in prefix list.  Should I2 limit the size of an RTBH to say something smaller than /24. 

1. Discussion: Sinn: recommends leaving it as in prefix range, as under duress this could cause confusion/delay if more specific RTBH is needed to stop traffic.  Their network will leverage blackholing as supported by upstream neighbors.

Contact Jeff if you have further thoughts on RTBH. Call ended 4:42 EDT