Home » MPLS Deep Dive: Label Switching, LSP, LDP, RSVP-TE, Traffic Engineering และ MPLS VPN
MPLS Deep Dive: Label Switching, LSP, LDP, RSVP-TE, Traffic Engineering และ MPLS VPN
MPLS Deep Dive: Label Switching, LSP, LDP, RSVP-TE, Traffic Engineering และ MPLS VPN
MPLS (Multiprotocol Label Switching) เป็น forwarding mechanism ที่ใช้ labels แทน IP lookup Label Switching ทำให้ forwarding เร็วขึ้น, LSP (Label Switched Path) คือเส้นทางที่ packets วิ่ง, LDP (Label Distribution Protocol) แจก labels อัตโนมัติ, RSVP-TE ทำ Traffic Engineering เลือกเส้นทางตามต้องการ, Traffic Engineering จัดสรร bandwidth อย่างมีประสิทธิภาพ และ MPLS VPN ให้บริการ VPN สำหรับลูกค้าหลายราย
แม้ SD-WAN กำลังมาแรง แต่ MPLS ยังเป็น backbone ของ ISP และ enterprise WAN ทั่วโลก: ISPs ทุกรายใช้ MPLS ใน core network, enterprise ใช้ MPLS VPN เชื่อมสาขา, mobile operators ใช้ MPLS ใน backhaul MPLS ให้ QoS guarantee, traffic engineering, และ multi-service capability ที่ SD-WAN (overlay) ยังทำไม่ได้ดีเท่า ความเข้าใจ MPLS จำเป็นสำหรับ network engineers ทุกคน
MPLS Label Operations
| Operation |
Where |
Action |
| Push (Impose) |
Ingress LER (Label Edge Router) |
Add label to packet — packet enters MPLS domain |
| Swap |
LSR (Label Switch Router) — transit |
Replace incoming label with outgoing label — forward based on label table |
| Pop (Dispose) |
Egress LER |
Remove label — packet exits MPLS domain back to IP forwarding |
| PHP (Penultimate Hop Popping) |
Second-to-last router |
Remove label one hop early → egress LER does normal IP lookup (reduces load) |
Label Distribution Protocols
| Protocol |
How |
Use Case |
| LDP |
Automatic label distribution — follows IGP shortest path |
Default MPLS label distribution — simple, no TE |
| RSVP-TE |
Signal explicit paths with bandwidth reservation |
Traffic engineering — choose path, reserve bandwidth |
| MP-BGP |
Distribute VPN labels (VPNv4/VPNv6 address families) |
MPLS L3 VPN — PE-to-PE label distribution |
| Segment Routing (SR) |
Source routing with segments — no LDP/RSVP needed |
Modern replacement for LDP/RSVP-TE — simpler, scalable |
MPLS Traffic Engineering
| Feature |
รายละเอียด |
| Problem |
IGP shortest path → some links overloaded, others underutilized |
| Solution |
RSVP-TE creates explicit LSPs → traffic follows engineered path (not just shortest) |
| CSPF |
Constrained Shortest Path First — find path with enough bandwidth, meeting constraints |
| Bandwidth Reservation |
Reserve bandwidth along path → guaranteed capacity for critical traffic |
| Fast Reroute (FRR) |
Pre-computed backup path → 50ms failover (vs seconds for IGP convergence) |
| Auto-Bandwidth |
Automatically adjust reserved bandwidth based on actual traffic measurements |
| Affinity/Coloring |
Tag links with colors → TE tunnels avoid/include specific link types |
MPLS L3 VPN
| Component |
Role |
Description |
| CE (Customer Edge) |
Customer router |
Connects to PE — runs routing protocol (BGP, OSPF, static) with PE |
| PE (Provider Edge) |
Service delivery |
Maintains VRF per customer — MP-BGP distributes VPN routes to other PEs |
| P (Provider) |
Core transit |
Label switching only — no customer route knowledge (scalable) |
| VRF |
Virtual Routing |
Per-customer routing table on PE — isolates customer A from customer B |
| RD (Route Distinguisher) |
Make routes unique |
Prefix + RD = unique VPNv4 route (even if customers use same IP ranges) |
| RT (Route Target) |
Control route distribution |
Export RT: advertise routes | Import RT: receive routes → flexible VPN topologies |
| Label Stack |
Two labels |
Outer = transport (LDP/RSVP) to egress PE | Inner = VPN (MP-BGP) to correct VRF |
MPLS L2 VPN
| Type |
How |
Use Case |
| VPWS (Virtual Private Wire Service) |
Point-to-point L2 circuit over MPLS → pseudowire |
Replace leased lines — L2 connectivity between 2 sites |
| VPLS (Virtual Private LAN Service) |
Multipoint L2 service → emulate LAN switch over MPLS |
Multi-site L2 connectivity — all sites in same broadcast domain |
| EVPN-MPLS |
EVPN control plane + MPLS data plane |
Modern replacement for VPLS — better MAC learning, multi-homing |
Segment Routing (Modern MPLS)
| Feature |
Traditional MPLS |
Segment Routing |
| Label Distribution |
LDP + RSVP-TE (separate protocols) |
IGP extensions (IS-IS/OSPF) — no LDP/RSVP needed |
| State |
Per-flow state on every router (RSVP-TE) |
Source routing — state only at ingress, transit = stateless |
| Traffic Engineering |
RSVP-TE with CSPF |
SR-TE with segment lists (Flex-Algo, SR Policy) |
| Scalability |
Limited by RSVP state on transit routers |
Highly scalable — no transit state |
| Fast Reroute |
FRR with backup tunnels |
TI-LFA (Topology Independent Loop-Free Alternate) — automatic |
| Data Plane |
MPLS labels |
SR-MPLS (labels) or SRv6 (IPv6 extension headers) |
ทิ้งท้าย: MPLS = Backbone of Service Provider Networks
MPLS Deep Dive Operations: push (ingress), swap (transit), pop (egress), PHP (penultimate hop) Label Distribution: LDP (auto, follows IGP), RSVP-TE (explicit path, bandwidth), MP-BGP (VPN labels) Traffic Engineering: CSPF path computation, bandwidth reservation, FRR (50ms failover), auto-bandwidth L3 VPN: CE-PE-P architecture, VRF isolation, RD (unique routes), RT (flexible distribution), 2-label stack L2 VPN: VPWS (point-to-point), VPLS (multipoint LAN), EVPN-MPLS (modern) Segment Routing: replaces LDP/RSVP-TE, source routing, stateless transit, TI-LFA, SR-MPLS/SRv6 Key: MPLS powers every ISP and large enterprise WAN — understanding it is essential for network engineering
อ่านเพิ่มเติมเกี่ยวกับ OSPF Deep Dive LSA Types Area Design Virtual Links และ BGP Deep Dive eBGP iBGP Path Attributes ที่ siamlancard.com หรือจาก icafeforex.com และ siam2r.com
