Interview Questions — Phase 3 · MPLS & L3VPN¶
These self-test questions target Google Network Infrastructure Engineer (NIE), Service Provider Backbone Engineer, and WAN Systems Engineer technical interview scenarios.
MPLS Underlay & LDP Mechanics¶
How does MPLS label switching differ from standard IP routing?
Standard IP routing performs a longest-prefix match (LPM) lookup on the 32-bit destination IPv4 address at every router hop. MPLS (Multiprotocol Label Switching) inspects a short, fixed-length 20-bit label in the MPLS shim header and performs an O(1) exact-match hardware table lookup to swap or pop the label, bypassing full routing table lookups at core P routers.
What is Penultimate Hop Popping (PHP) and why is it used?
Penultimate Hop Popping (PHP) is an MPLS optimization specified in RFC 3032. The egress PE advertises an Implicit Null Label (Label 3) to its upstream P router. The upstream P router pops the transport label before forwarding the packet to the egress PE. This prevents the egress PE from having to perform two hardware label lookups (transport label + service label).
L3VPN & MP-BGP (RFC 4364)¶
What is the difference between a Route Distinguisher (RD) and a Route Target (RT)?
- Route Distinguisher (RD — 64 bits): Makes overlapping IPv4 customer addresses globally unique by prepending an RD to form a 96-bit VPNv4 prefix (
RD + IPv4 = VPNv4). - Route Target (RT — Extended Community): Defines VRF routing policy. Determines which VRFs import and export specific VPNv4 routes across the provider network.
How does a PE router differentiate between traffic belonging to different customer VRFs upon receiving a packet from the core?
The packet carries an Inner VPN Service Label in its MPLS label stack. When MP-BGP advertises a VPNv4 route, it allocates a unique service label for that prefix/VRF. The egress PE looks up the inner label in its LFIB (Label Forwarding Information Base) to identify the destination VRF and egress interface.
Inter-AS L3VPN Options A, B, and C¶
Compare Inter-AS L3VPN Options A, B, and C in terms of ASBR scalability.
- Option A (Back-to-Back VRFs): Low scalability. ASBRs must configure sub-interfaces, VRFs, and individual eBGP sessions for every customer. High memory/CPU overhead.
- Option B (Inter-AS MP-eBGP VPNv4): Medium/High scalability. ASBRs exchange VPNv4 routes directly over MP-eBGP and rewrite Next-Hops/labels. No customer VRFs required on ASBRs.
- Option C (BGP-LU RFC 3107 + Multi-hop MP-eBGP): Hyperscale. ASBRs carry zero customer VPN routes, only exchanging PE loopback reachability via BGP Labeled Unicast (BGP-LU). PEs peer directly via multi-hop MP-eBGP.
Why do hyperscalers (Google, AWS, Meta) prefer Inter-AS Option C over Option A or B?
Option C completely decouples the backbone ASBR core from customer VPN control plane state. ASBRs only store loopback prefixes (\sim 10,000 PEs) instead of millions of customer VPN routes. This allows scaling network capacity infinitely without exhausting ASBR router memory or TCAM state.