Switching, Routing, and Bridging Terminology

After discussing networking layers and addressing, it’s time to focus on moving packets across a network. Vendors love to use ill-defined terms like switching instead of forwarding, routing, or bridging, so let’s start with the terminology.

Connecting all relevant devices to a single cable would indubitably simplify any networking stack, but unfortunately, we’re almost never that lucky. We need devices in the network (typically with multiple interfaces) that perform packet forwarding between end nodes.

read more add comment

Worth Reading: BGP Unnumbered in 2025

Gabriel sent me a pointer to a blog post by Rudolph Bott describing the details of BGP Unnumbered implementations on Nokia, Juniper, and Bird.

Even more interestingly, Rudolph points out the elephant I completely missed: RFC 8950 refers to RFC 2545, which requires a GUA IPv6 next hop in BGP updates (well, it uses the SHALL wording, which usually means “troubles ahead”). What do you do if you’re running EBGP on an interface with no global IPv6 addresses? As expected, vendors do different things, resulting in another fun interoperability exercise.

Finally, there’s RFC 7404 that advocates LLA-only infrastructure links, so we might find the answer there. Nope; it doesn’t even acknowledge the problem in the Caveats section.

For even more information, read the Unnumbered IPv4 Interfaces and BGP in Data Center Fabrics blog posts.

see 2 comments

How ARP Killed a Static Route

The amount of weird stuff we discover in netlab integration tests is astounding, or maybe I have a knack for looking into the wrong dark corners (my wife would definitely agree with that). Today’s special: when having two next hops kills a static route.

TL&DR: default ARP settings on a multi-subnet Linux host are less than optimal.

We use these principles when creating netlab integration tests:

  • They should contain a single device-under-test and a bunch of attached probes.
  • They should test a single feature.
  • They should not rely on the device-under-test. All validation has to be done on probes.

How do you test static routes under these restrictions? Here’s what we did:

read more see 1 comments

Rant: You Should Have Written a Book

I apologize for the rant; I have to vent my frustration with people whose quantity of opinions seems to be exceeding their experience (or maybe they’re coming from an alternate universe with different laws of physics, which would be way cool but also unlikely). You’ve been warned; please feel free to move on or skip the rant part of the blog post.

Rant mode: ON

This is the (unedited) gem I received after making some of my EVPN videos public:

read more add comment

Redundant Small Site Multihoming

The original Small Site Multihoming article has generated many responses, most of them being questions about the redundant implementation of the same principles. In this article, we’ll thus add a second customer router to add redundancy to the small site multi-homing design. The final design will still retain the administrative simplicity of the original solution – with no need to own public IP address space, autonomous system number, or to run Border Gateway Protocol (BGP).

read more add comment

Response: NAT Traversal Mess

Let’s look at another part of the lengthy comment Bob left after listening to the Rise of NAT podcast. This one is focused on the NAT traversal mess:

You mentioned that only video-conferencing and BitTorrent use client-to-client connectivity (and they are indeed the main use cases), but hell, do they need to engineer complex systems to circumvent these NATs and firewalls: STUN, TURN, ICE, DHT…

Cleaning up the acronym list first: DHT is unlike the others and has nothing to do with NAT.

read more see 2 comments

iBGP Local-AS Details

Did you know you could use the neighbor local-as BGP functionality to fake an iBGP session between different autonomous systems? I knew Cisco IOS supported that monstrosity for ages (supposedly “to merge two ISPs that have different AS numbers”) and added the appropriate tweaks1 into netlab when I added the BGP local-as support in release 1.3.1. Someone couldn’t resist pushing us down that slippery slope, and we ended with IBGP local-as implemented on 18 platforms (almost a dozen network operating systems).

I even wrote a related integration test, and all our implementations passed it until I asked myself a simple question: “But does it work?” and the number of correct implementations that passed the test without warnings dropped to zero.

read more add comment

Transitioning into Networking, 2025 Edition

Elmer sent me the following question:

I’ve been working in systems engineering (Linux, virtualization, infrastructure ops) and am considering shifting toward network engineering or architecture. I got my CCNA years ago and started CCNP but didn’t continue.

I’d really appreciate any thoughts you might have on how someone with my background could best make that transition today, especially with how things are evolving around automation and the cloud.

I keep answering a variant of this question every other year or so (2019, 2021, 2023, 2024). I guess it’s time for another answer, so here we go.

read more see 1 comments

ARP Challenges in EVPN/VXLAN Symmetric IRB

Whenever I claimed that EVPN is The SIP of Networking, vendor engineers quickly told me that “EVPN interoperability is a solved problem” and that they run regular multi-vendor interoperability labs to iron out the quirks. As it turns out, things aren’t as rosy in real life; it’s still helpful to have an EVPN equivalent of the DTMF tone generators handy.

I encountered a particularly nasty quirk when running the netlab EVPN integration test using symmetric IRB with an anycast gateway between Nokia SR Linux (or Juniper vSwitch) and FRR container.

Lab topology

Lab topology

read more add comment

Comparing IP and CLNP: Local (Node) Multihoming

Another area where CLNP is a clear winner when compared to the TCP/IP stack is multi-homed nodes (nodes with multiple interfaces, not site multi-homing, where whole networks are connected to two upstream providers).

Multi-homed TCP/IP nodes must have multiple IP addresses because IP uses address interfaces. There is no well-defined procedure in TCP/IP for how a multi-homed node should behave. In the early days of TCP/IP, they tried to address that in RFC 1122 (Host Requirements RFC), but even then, there were two ideas about dealing with multiple interfaces: the strong and weak end system models (more details).

read more see 1 comments
Sidebar