Segment Routing: Managing SRv6 Header Overhead with uSID and SRm6

In my last post, we looked at how to configure SRv6 on the Nokia 7750 SR. One big challenge with SRv6 is the header overhead. Since every "Segment Identifier" (SID) is a full 128-bit IPv6 address inside the SRH header, the header gets very large very quickly.

This extra weight often causes fragmentation. When a packet is too big for the MTU of a link, it has to be broken into pieces.

This is bad for two reasons:

1. It slows down the network because of the processing power needed for fragmentation/reassembly.

2. It's a security risk, as hackers often use fragmentation attacks to sneak past security filters.

We have two main options to make the IPv6 headers smaller:

1. Micro-SID (uSID)

Micro-SID is strongly recommended by Nokia and Cisco. Instead of giving every hop a full 128-bit address, uSID uses a "container" concept.

How it works: It takes one 128-bit address and divides it up. The first part (the Locator) identifies the network, and the rest of the bits are used for "mini-IDs" (16-bit length in FP-based platforms or 32-bit for other platforms in Nokia).

Ideally, 6 mini-SID hops are injected inside a single 128-bit address. This keeps the header small and prevents the packet from needing to be fragmented.

2. SRm6

This is a different approach, mostly recommended by Juniper. Juniper always does things differently!

SRm6 runs Segment Routing over "Native IPv6" without using the heavy SRv6 header. It uses much smaller IDs—only 16 or 32 bits per destination. This makes the packet look more like a standard IPv6 packet, which saves a lot of space.

---

Lab Example: SRH Header Capture

Here is a screenshot from my lab. You can see the SRH (Segment Routing Header) in the packet capture. Notice how much space the segment list takes up when we use standard 128-bit SIDs.

---

Final Thoughts

In my view, if you are working in a multi-vendor environment, sticking with uSID is the better path for SRv6. It’s becoming the industry standard for keeping the data plane efficient while ensuring compatibility between the big players like Nokia and Cisco.