SRv6: LCM vs GCM

SRv6: LCM vs GCM — What Telcos Really Need to Know

✍️ Written by: RJS Expert
A definitive guide to understanding LCM and GCM in SRv6 uSID architecture and congestion management.

As SRv6 + uSID goes mainstream across Telcos, two terms create constant confusion in architecture discussions:

LCM and GCM

Both are critical — and both have two different meanings depending on context.

Here's the cleanest way to understand them 👇

1️⃣ In SRv6 uSID Architecture

🔵 LCM — Locator Compressed Mode (Ingress Encoding)

Defines how SRv6 paths are encoded at the policy headend using compressed uSID containers.

👨‍💼 Operators:

• Design locator/uSID plan
• Choose compression model
• Build SRv6 TE policies
• Integrate with SDN controllers

🏭 Vendors:

• Implement uSID headend behaviors
• Insert SRH with compressed containers
• Support BGP-LS/PCEP/YANG for TE

👉 LCM = How the path is encoded.

🟢 GCM — Generalized Compressed Mode (Transit Forwarding)

Defines how intermediate routers forward uSID packets using shift + rewrite operations.

👨‍💼 Operators:

• Ensure all nodes support GCM
• Validate ASIC capability
• Enable SRv6 forwarding domain-wide

🏭 Vendors:

• Implement End.uSID / T.uSID behaviors
• Deliver line-rate uSID forwarding
• Provide telemetry for SRv6 operations

👉 GCM = How the packet moves.

2️⃣ In Congestion Management

🔵 LCM — Local Congestion Management

Node-level stability:

• Queueing (PQ/WFQ/WRR)
• WRED/ECN
• Buffer & burst control
• TI-LFA protection

👉 Keeps individual routers healthy.

🟢 GCM — Global Congestion Management

Network-wide intelligence using telemetry + controllers:

• Detect congestion end-to-end
• Re-optimize SRv6 TE policies
• Predict hotspots via AI/ML
• Redirect flows for SLA guarantees

👉 Optimizes the entire network, not just a node.

📊 Quick Reference Matrix

Context	LCM	GCM
SRv6 uSID	Locator Compressed Mode Path encoding at headend	Generalized Compressed Mode Transit forwarding behavior
Congestion Mgmt	Local Congestion Management Node-level QoS & stability	Global Congestion Management Network-wide optimization

🎯 The Bottom Line

✔ LCM (uSID)	= Path Encoding
✔ GCM (uSID)	= Forwarding Behavior
✔ LCM (Congestion)	= Local Node Control
✔ GCM (Congestion)	= Global Network Optimization

Understanding both perspectives is essential for SRv6 success.
Operators design the SID architecture and QoS strategy.
Vendors must deliver silicon, OS, and telemetry that make it real.

🔍 Deep Dive: SRv6 uSID Architecture

LCM: Locator Compressed Mode Details

What happens at the headend:

1. Controller (or local policy) computes the desired SRv6 path
2. Headend router encodes multiple uSIDs into a compressed SID container
3. SRH (Segment Routing Header) is inserted with the compressed list
4. Packet enters the SRv6 domain with full path information

Key Design Decisions:

• uSID block size: Typically /48 or /32 locator with 16-bit uSIDs
• Compression ratio: How many uSIDs fit in one IPv6 address (typically 6-8)
• Policy source: BGP-LS, PCEP, local config, or SDN controller
• Fallback behavior: What happens if uSID depth exceeds container capacity

GCM: Generalized Compressed Mode Details

What happens at transit nodes:

1. Router receives packet with compressed uSID list
2. Inspects the active uSID (next segment to process)
3. Executes local SID behavior (End.uSID, T.uSID, etc.)
4. Shifts the uSID container to expose the next segment
5. Forwards to the next hop without additional encapsulation

Critical Requirements:

• Hardware support: ASIC must support SRv6 uSID shift operations at line rate
• Consistent uSID block: All nodes must use the same locator/uSID structure
• Behavior consistency: End.uSID must behave identically across all vendors
• Telemetry integration: Track uSID forwarding metrics for troubleshooting

🚦 Congestion Management: Local vs Global

Aspect	LCM (Local)	GCM (Global)
Scope	Single node	Entire network domain
Decision Point	Router's local QoS engine	Centralized controller/analytics
Reaction Time	Microseconds	Seconds to minutes
Input Data	Queue depth, buffer state	Telemetry, flow data, topology
Action Taken	Drop, mark, delay packets	Reroute flows, adjust TE policies
Technology	WRED, ECN, shaping, policing	PCEP, BGP-LS, SDN, AI/ML
Goal	Prevent local overload	Optimize network-wide performance

💡 Pro Tip:

You need both LCM and GCM for modern telco networks:

• LCM protects individual routers from collapse during microbursts
• GCM prevents congestion from happening in the first place by intelligent path selection
• Without LCM, nodes fail during transient overload
• Without GCM, you're always reactive, never proactive

🛠️ Implementation Considerations

For Network Operators:

1. Design Phase:
   • Define uSID block allocation strategy (LCM)
   • Choose compression model (LCM)
   • Validate ASIC support for uSID forwarding (GCM)
   • Plan QoS policy at node level (LCM)
   • Architect telemetry and SDN controller integration (GCM)
2. Testing Phase:
   • Verify uSID encoding at headend (LCM)
   • Confirm shift operations at transit nodes (GCM)
   • Load test local congestion controls (LCM)
   • Validate end-to-end path optimization (GCM)
3. Operations Phase:
   • Monitor uSID policy hit rates
   • Track congestion events (local vs global)
   • Tune AI/ML models for predictive rerouting
   • Maintain consistency across multi-vendor environments

"SRv6 uSID isn't just a feature — it's a transport transformation."

Understanding LCM and GCM in both contexts is the foundation of that transformation.

Final Word:

The confusion around LCM and GCM isn't accidental — it reflects the dual nature of modern telco networks:

• One layer encodes and forwards packets (SRv6 uSID)
• Another layer manages stability and performance (congestion management)

Mastering both is what separates architectural clarity from operational chaos.