Home » Network Traffic Engineering: MPLS-TE, RSVP-TE, Bandwidth Reservation, Auto-Bandwidth และ PCE
Network Traffic Engineering: MPLS-TE, RSVP-TE, Bandwidth Reservation, Auto-Bandwidth และ PCE
Network Traffic Engineering: MPLS-TE, RSVP-TE, Bandwidth Reservation, Auto-Bandwidth และ PCE
Network Traffic Engineering ควบคุมเส้นทางของ traffic ให้ใช้ทรัพยากร network อย่างมีประสิทธิภาพ MPLS-TE สร้าง label-switched paths ที่กำหนดเส้นทาง, RSVP-TE เป็น signaling protocol จอง bandwidth, Bandwidth Reservation รับประกันแบนด์วิดท์, Auto-Bandwidth ปรับ bandwidth reservation อัตโนมัติ และ PCE (Path Computation Element) คำนวณเส้นทางที่ดีที่สุดจากมุมมองทั้ง network
Traffic Engineering แก้ปัญหาที่ IGP shortest path ทำไม่ได้: IGP (OSPF/IS-IS) เลือกเส้นทางสั้นที่สุดเสมอ → ถ้า shortest path เต็ม แต่ longer path ว่าง → IGP ไม่ย้าย traffic ไปได้ TE ให้ control path ได้เต็มที่: กำหนดเส้นทาง, จอง bandwidth, หลีกเลี่ยง congested links, ทำ fast-reroute ISPs ใช้ TE เพื่อรับประกัน SLA ให้ลูกค้า enterprise และป้องกัน congestion ในช่วง peak hours
MPLS-TE Concepts
| Concept |
Description |
| LSP (Label Switched Path) |
Pre-established path through network — traffic follows this path regardless of IGP shortest path |
| Head-End |
Ingress router that creates and maintains the TE tunnel — source of the LSP |
| Tail-End |
Egress router — destination of the TE tunnel |
| TE Tunnel |
Logical interface on head-end representing the LSP — route traffic into tunnel |
| Explicit Path |
Administrator specifies exact hops the LSP must traverse — full control over path |
| CSPF |
Constrained Shortest Path First — compute path considering constraints (bandwidth, affinity, color) |
RSVP-TE Signaling
| Message |
Direction |
Function |
| Path |
Head-End → Tail-End |
Sent hop-by-hop toward tail-end — carries ERO (Explicit Route Object) defining the path |
| Resv |
Tail-End → Head-End |
Sent back along the path — reserves bandwidth and assigns labels at each hop |
| PathTear |
Head-End → Tail-End |
Tear down the LSP — release reserved bandwidth and labels |
| ResvTear |
Tail-End → Head-End |
Release reservation in reverse direction |
| PathErr |
Toward Head-End |
Error notification — path computation failed, admission control rejected |
| ResvErr |
Toward Tail-End |
Reservation error — bandwidth not available at some hop |
| Hello |
Between neighbors |
Detect neighbor failures for fast-reroute — keepalive mechanism |
Bandwidth Reservation & Admission Control
| Feature |
Detail |
| Reservable BW |
Each link advertises available bandwidth via IGP-TE extensions (OSPF-TE/IS-IS-TE) |
| Admission Control |
Before establishing LSP: check if requested bandwidth is available on every hop → reject if not |
| Preemption |
Higher priority LSP can preempt (kick out) lower priority LSP to claim bandwidth |
| Setup Priority |
0 (highest) to 7 (lowest) — priority for establishing new LSP |
| Hold Priority |
0 (highest) to 7 (lowest) — priority for keeping established LSP (resist preemption) |
| Sub-Pool |
Reserve portion of link bandwidth for specific traffic class (e.g., voice vs data) |
Auto-Bandwidth
| Feature |
Detail |
| Problem |
Static bandwidth reservation: too high = waste, too low = congestion |
| Solution |
Auto-bandwidth monitors actual traffic → periodically adjusts reservation to match real usage |
| Sample Interval |
Measure actual throughput every 5 minutes (configurable) |
| Adjustment |
Every adjustment interval (e.g., 24 hours): if measured > reserved → increase | if measured < reserved → decrease |
| Min/Max |
Set minimum and maximum bandwidth limits — prevent under/over reservation |
| Overflow |
If traffic exceeds reserved by threshold → trigger immediate re-optimization (don’t wait for interval) |
PCE (Path Computation Element)
| Feature |
Detail |
| What |
Centralized server that computes optimal paths for TE tunnels — has global view of network topology and bandwidth |
| Why |
Head-end router only has local TED (Traffic Engineering Database) → may not find globally optimal path |
| PCEP |
PCE Communication Protocol — routers (PCC) request path computation from PCE server |
| Stateful PCE |
PCE knows all active LSPs → compute paths considering existing reservations → global optimization |
| Active PCE |
PCE proactively re-optimizes LSPs → move LSPs to better paths when network changes |
| Multi-Domain |
PCE can compute paths across multiple IGP areas/AS → inter-domain TE |
| SR-PCE |
PCE for Segment Routing — compute segment lists instead of RSVP-TE signaled paths |
TE Fast-Reroute (FRR)
| Type |
Protection |
How |
| Facility Backup (Bypass) |
Link protection |
Pre-compute backup tunnel around protected link → on failure, switch to bypass in < 50ms |
| One-to-One Backup |
Per-LSP protection |
Each protected LSP has dedicated backup path → more resource intensive but more flexible |
| Node Protection |
Protect against node failure |
Bypass tunnel goes around protected node (not just link) → protects against router failure |
| Bandwidth Protection |
Guarantee BW on backup path |
Reserve bandwidth on backup path → guaranteed QoS even during failure (at cost of reserved BW) |
ทิ้งท้าย: Traffic Engineering = Control Where Traffic Flows
Network Traffic Engineering MPLS-TE: create explicit LSPs that override IGP shortest path → utilize all links efficiently RSVP-TE: signaling protocol — Path message (forward), Resv message (backward reserve BW + assign labels) Bandwidth: admission control, preemption (priority 0-7), sub-pool for traffic classes Auto-BW: monitor actual traffic → adjust reservation automatically — prevents waste and congestion PCE: centralized path computation — global view → optimal paths, stateful/active, multi-domain, SR-PCE FRR: facility bypass (link protection), node protection, < 50ms failover — pre-computed backup paths Key: TE gives operators control over traffic placement — essential for SLAs, congestion avoidance, link utilization
อ่านเพิ่มเติมเกี่ยวกับ MPLS Deep Dive Label Switching LSP LDP RSVP-TE และ Segment Routing SR-MPLS SRv6 SRGB Node-SID TI-LFA ที่ siamlancard.com หรือจาก icafeforex.com และ siam2r.com
