Customer routing in a provider network: why VRFs
A service provider core carries traffic for many customers over the same routers. Each customer expects private routing, overlapping addresses allowed and no leakage. This chapter explains the picture and why virtual routing and forwarding (VRF) is the tool.
Analogy: one apartment block, many private flats
Many families share one building, but each has a locked flat with its own post box. A VRF is a locked flat for routes: one router, many separate routing tables.
The roles
| Device | Role |
|---|---|
| CE (customer edge) | The customer's router at the site |
| PE (provider edge) | Provider router that faces customers and holds VRFs |
| P (provider) | Core router that only moves labelled traffic |
What VRFs give
- Separate routing tables per customer on the same PE.
- Overlapping address ranges between customers without conflict.
- Per-customer policy and control over what is exchanged.
- A clean base for MPLS VPN services that follow in later modules.
What this module covers
PE-CE routing with static routes, an IGP and eBGP; overlapping addresses; controlled sharing between VRFs; verification and troubleshooting. The core and labels come later.
Customers connect to PEs in VRFs.
Two customers both use 192.0.2.0/24 internally. The PE keeps their routes in two VRFs, CUST-A and CUST-B, so each customer sees only its own 192.0.2.0/24 and nothing collides.
Common mistake: putting two customers in the global routing table. Overlapping ranges and leaks follow.
Same range, two customers
A new customer used the same internal range as an existing one. Because both were in separate VRFs, no change was needed on the provider side.
Lesson: separate tables solve overlap.
Why use VRFs on a provider edge router?
They give each customer a separate routing table on the same router, allowing overlapping addresses and isolation.
Key takeaways
- CE, PE and P roles.
- VRF = separate table.
- Overlap allowed.
- Isolation by default.
- Base for MPLS VPNs.