Bruce Davie continues documenting the tradeoffs we had to make in networking, this time with Was MPLS Traffic Engineering Worthwhile? I found this bit particularly familiar:

It wasn’t hard to make a theoretical argument that MPLS-TE could improve network performance and average link utilization, by moving traffic from congested links to uncongested ones. The hard part was proving that it would actually do a better job in practice than the more traditional methods such as using link weights and multipath routing to achieve the same ends.

One of my readers successfully deployed LDPv6 in their production network:

We are using LDPv6 since we started using MPLS with IPv6 because I was used to OSPF/OSPFv3 in dual-stack deployments, and it simply worked.

Not everyone seems to be sharing his enthusiasm:

Now some consultants tell me that they know no-one else that is using LDPv6. According to them “everyone” is using 6PE and the future of LDPv6 is not certain.

Henk Smit left a wonderful comment discussing various scalability aspects of SR-MPLS. Let’s go through the points he made:

When you have a thousand routers in your networks, you can put all of them in one (IS-IS) area. Maybe with 2k routers as well. But when you have several thousand routers, you want to use areas, if only to limit the blast-radius.

Absolutely agree, and as RFC 3439 explained in more eloquent terms than I ever could:

Whenever I compare MPLS-based Segment Routing (SR-MPLS) with it’s distant IPv6-based cousin (SRv6), someone invariably mentions the specter of large label stacks that some hardware cannot handle, for example:

Do you think vendors current supported label max stack might be an issue when trying to route a packet from source using Adj-SIDs on relatively big sized (and meshed) cores? Many seem to be proposing to use SRv6 to overcome this.

I’d dare to guess that more hardware supports MPLS with decent label stacks than SRv6, and if I’ve learned anything from my chats with Laurent Vanbever, it’s that it sometimes takes surprisingly little to push the traffic into the right direction. You do need a controller that can figure out what that little push is and where to apply it though.

Most networking engineers immediately think about VXLAN and data center switches when they hear about EVPN. While that’s the most hyped use case, EVPN standardization started in 2012 as a layer-2 VPN solution on top of MPLS transport trying to merge the best of VPLS and MPLS/VPN worlds.

If you want to understand how any technology works, and what its quirks are, you have to know how it was designed to be used. In this blog post we’ll start that journey exploring the basics of EVPN used in a simple MLPS network with three PE-routers:

Here’s an interesting question randomly appearing in my Twitter feed:

If you had a greenfield network, would you choose SR-MPLS, or SRv6? And why?

TL&DR: SR-MPLS, assuming you’re building a network providing end-to-end connectivity between hardware edge devices.

Now for the why part of the question:

One of my readers asked for my opinion about this question…

… and I promised something longer than 280 characters.

Jeff Tantsura left me tantalizing hint after reading the BGP Labeled Unicast on Cisco IOS blog post:

Read carefully “Relationship between SAFI-4 and SAFI-1 Routes” section in RFC 8277

The start of that section doesn’t look promising (and it gets worse):

It is possible that a BGP speaker will receive both a SAFI-1¹ route for prefix P and a SAFI-4² route for prefix P. Different implementations treat this situation in different ways.

Now for the details:

netlab release 1.2.0 adds full-blown MPLS and MPLS/VPN support:

• VRF definitions and layer-3 VRFs
• VRF-aware OSPF, IS-IS and BGP
• Traditional MPLS with LDP (SR-MPLS was already available)
• BGP Labeled Unicast
• MPLS/VPN: VPNv4 and VPNv6 address family support

It’s never been easier to build full-blown MPLS/VPN labs ;)… if you’re OK with using Cisco IOS or Arista EOS. Please feel free to submit a PR to add support for other platforms.

You might want to start with the VRF tutorial to see how simple it is to define VRFs, and follow the installation guide to set up your lab – if you’re semi-fluent in Linux (and don’t care about data plane quirks), the easiest option would be to run Arista cEOS.

A week ago I described how Cisco IOS implemented BGP Labeled Unicast. In this blog post we’ll focus on Arista EOS using the same lab as before:

While researching the BGP RFCs for the Three Dimensions of BGP Address Family Nerd Knobs, I figured out that the BGP Labeled Unicast (BGP-LU, advertising MPLS labels together with BGP prefixes) uses a different address family. So far so good.

Now for the intricate bit: a BGP router might negotiate IPv4 and IPv4-LU address families with a neighbor. Does that mean that it’s advertising every IPv4 prefix twice, once without a label, and once with a label? Should that be the case, how are those prefixes originated and how are they stored in the BGP table?

As always, the correct answer is “it depends”, this time on the network operating system implementation. This blog post describes Cisco IOS behavior, a follow-up one will focus on Arista EOS.

Hannes Gredler wrote an interesting comment to my Segment Routing vs LDP in Hub-and-Spoke Networks blog post:

In 2014 when I did the first prototype implementation of MPLS-SR node labels, I was stunned that just with an incremental add of 500 lines of code to the vanilla IPv4/IPv6 IS-IS codebase I got full any-to-any connectivity, no sync issues, no targeted sessions for R-LFA …. essentially labeled transport comes for free.

Based on that, one has to wonder “why did we take the LDP detour and all the complexity it brings?”. Here’s what Hannes found out:

In the Segment Routing vs LDP in Hub-and-Spoke Networks blog post I explained why you could get into interesting scaling issues when running MPLS with LDP in a large hub-and-spoke network, and how you can use Segment Routing (MPLS edition) to simplify your design.

Now imagine you’d like to offer VPLS services between hubs and spokes, and happen to be using equipment that uses targeted LDP sessions to signal pseudowires. Guess what happens next…

I got an interesting question that nicely illustrates why Segment Routing (the MPLS variant) is so much better than LDP. Imagine a redundant hub-and-spoke network with hundreds of spokes. Let’s settle on 500 spokes – IS-IS supposedly has no problem dealing with a link-state topology of that size.

Let’s further assume that all routers advertise only their loopbacks¹ and that we’re using unnumbered hub-to-spoke links to minimize the routing table size. The global routing table thus contains ~500 entries. MPLS forwarding tables (LFIB) contain approximately as many entries as each router assigns a label to every prefix in the routing table². What about the LDP table (LIB – Label Information Base)?

Last week I published a link to Pete Crocker’s RSVP-TE lab, but there’s more: he created another lab using the same topology that uses SR-MPLS with IS-IS to get the job done.

Jeroen Van Bemmel did something similar for SR Linux: his lab topology has fewer devices (plus SR Linux runs in containers), so it’s easily deployable on machines without humongous amount of memory.