Struggling to build a lab for IS-IS and Segment Routing without melting your laptop? I break down three distinct ways to get your topology running, from lightweight Docker containers to full-scale vendor emulation.

Mastering the Lab: IS-IS with Segment Routing (Without Melting Your CPU)

I still remember the first time I tried to simulate a carrier-grade MPLS network on my personal laptop. It was 2015, I was running GNS3, and I tried to spin up seven instances of Cisco IOS-XR. My fan sounded like a jet engine preparing for takeoff, and the room temperature rose by about ten degrees. Five minutes later? Blue Screen of Death.

I saw a thread recently where someone was begging for a lab topology to configure IS-IS with Segment Routing (SR). I felt that pain immediately. You want to learn the cool new tech—getting rid of LDP, simplifying traffic engineering—but the barrier to entry isn’t the configuration syntax; it’s the sheer weight of the virtual routers required to run it.

If you are stuck trying to find a lab environment that actually works, you are not alone. Here is how we tackle this at TechResolve.

The “Why”: It’s Not You, It’s The Bloat

The root cause of your frustration is that traditional router images (like XRv9k or vMX) are resource hogs. They expect to be running on heavy iron. When you try to cram an IS-IS SR topology (which needs a minimum of 3-4 nodes to be interesting) onto a standard workstation, you hit resource exhaustion before you even type router isis 1.

Furthermore, standard IOS images often don’t support Segment Routing. You need the “Service Provider” flavors (IOS-XE, IOS-XR, Junos), and obtaining those images legally for a home lab can be a navigational nightmare of licensing portals.

Solution 1: The Modern DevOps Way (Containerlab + FRR)

This is my go-to for 90% of my architecture modeling now. Forget heavy VMs. We use Containerlab running FRRouting (FRR). FRR is a Linux-based routing stack that supports IS-IS and Segment Routing natively. It boots in seconds and uses megabytes of RAM, not gigabytes.

Pro Tip: This approach is “platform agnostic.” You learn the protocol mechanics (Adjacency SIDs, Prefix SIDs) without getting bogged down in proprietary CLI syntax.

Here is a basic topology.yml to get you three nodes connected in a ring:

name: isis-sr-lab
topology:
  nodes:
    r1:
      kind: linux
      image: frrouting/frr:latest
    r2:
      kind: linux
      image: frrouting/frr:latest
    r3:
      kind: linux
      image: frrouting/frr:latest
  links:
    - endpoints: ["r1:eth1", "r2:eth1"]
    - endpoints: ["r2:eth2", "r3:eth2"]
    - endpoints: ["r3:eth3", "r1:eth3"]

Once inside the container (docker exec -it clab-isis-sr-lab-r1 vtysh), the config is shockingly simple:

router isis 1
 is-type level-2-only
 net 49.0001.0000.0000.0001.00
 segment-routing on
 segment-routing global-block 16000 23999
 segment-routing prefix 1.1.1.1/32 index 10 no-php-flag
!
interface eth1
 ip router isis 1
 ipv6 router isis 1

Solution 2: The “Classic” Heavy Lifter (EVE-NG + CSR1000v)

Sometimes you can’t escape the vendor CLI. If you are prepping for a CCIE Service Provider or validating a config for prod-core-01, you need the real deal. For this, I recommend EVE-NG over GNS3 simply for the stability of the backend, specifically using Cisco CSR1000v images (newer IOS-XE).

However, you have to be smart about resources. Don’t build a full mesh. Build a “kite” topology.

It’s hacky to starve the core routers of RAM, but as long as you aren’t loading full internet routing tables, the control plane will survive long enough for you to verify the SRGB (Segment Routing Global Block).

Solution 3: The ‘Nuclear’ Option (Cisco DevNet Sandbox)

If your laptop is an ultrabook and you don’t have a homelab server, stop trying to run this locally. You will only frustrate yourself.

The “Nuclear” option is to offload the compute entirely. I frequently tell my juniors to use the Cisco DevNet Sandbox. Specifically, look for the “IOS XR Programmability” sandboxes.

• Pros: It’s free. It runs on real hardware or high-end virtual instances. It’s already cabled up.
• Cons: You have to reserve a time slot, and you lose your config when the reservation ends.

It feels like “cheating” because you aren’t building the lab yourself, but remember: Your goal is to learn IS-IS Segment Routing, not how to troubleshoot QEMU CPU flags. If the infrastructure is blocking your learning, outsource the infrastructure.

Darian’s Final Thought: Start with Solution 1 (FRR). It forces you to understand the RFCs rather than just memorizing commands. Once you get the logic of Node SIDs and Adjacency SIDs down, the Cisco or Juniper syntax is just translation.

Darian Vance

Lead Cloud Architect & DevOps Strategist

With over 12 years in system architecture and automation, Darian specializes in simplifying complex cloud infrastructures. An advocate for open-source solutions, he founded TechResolve to provide engineers with actionable, battle-tested troubleshooting guides and robust software alternatives.

🤖 Frequently Asked Questions