Case Study: Deploy 50-Site MPLS Network untuk ISP Regional

Latar Belakang

ISP regional di Sumatera mengelola 50 Point of Presence (POP) tersebar di 10 provinsi. Jaringan lama menggunakan mesh VPN IPsec yang menyebabkan:

• Konfigurasi manual per site — error-prone, tidak konsisten
• Troubleshooting sulit — tidak ada visibility end-to-end
• Scalability — menambah site butuh hari, bukan jam
• QoS tidak terjamin — bandwidth sharing tidak adil antar pelanggan

Arsitektur Solusi

Layer 1: Core (Backbone)

• 2x MikroTik CRS326-24G-2S+ sebagai core switch/router di DC Jakarta & DC Medan
• OSPF Area 0 dengan authentication, fast hello (1s)
• MPLS LDP untuk label distribution
• RSVP-TE untuk explicit path engineering

Layer 2: Aggregation (Regional)

• CCR2004-16G-2S+ per wilayah (Sumatera Utara, Barat, Selatan, dll)
• OSPF Area 1-10 (per region), stub area untuk efisiensi
• Route redistribution OSPF ↔ BGP di border

Layer 3: Access (Site/POP)

• RB4011iGS+ / RB5009UG+S+IN per site
• Static route ke aggregation (simple, scalable)
• VRF per customer (pisah traffic enterprise vs residential)

Implementasi Kunci

MPLS L3VPN dengan BGP VPNv4

# Core: MPLS LDP + OSPF
/mpls ldp interface add interface=ether1-core
/routing ospf instance set default mpls-te=yes

# VRF Customer
/routing vrf add name=Customer-A routing-mark=customer-a
/routing bgp vrf add vrf=Customer-A rd=65001:100

# BGP VPNv4
/routing bgp peer add name=Core-BGP remote-address=10.0.0.2 remote-as=65001
/routing bgp peer add name=Core-BGP address-families=vpnv4

Traffic Engineering (RSVP-TE)

# Explicit path untuk avoid single point of failure
/mpls traffic-eng path add name=Primary-Path
/mpls traffic-eng path add name=Backup-Path

# Auto-bandwidth adjustment
/mpls traffic-eng tunnel add name=TE-Core-Aggregation auto-bandwidth=yes

QoS HTB per Paket Langganan

# Parent queue di interface WAN
/queue tree add name=Parent-WAN parent=ether1-WAN max-limit=1G

# Child queue per paket (misal: 50Mbps, 100Mbps, 200Mbps)
/queue tree add name=Pkt-50M parent=Parent-WAN max-limit=50M priority=2
/queue tree add name=Pkt-100M parent=Parent-WAN max-limit=100M priority=3
/queue tree add name=Pkt-200M parent=Parent-WAN max-limit=200M priority=4

# PCQ untuk fairness dalam satu paket
/queue type add name=PCQ-Download kind=pcq pcq-rate=0 pcq-classifier=dst-address
/queue type add name=PCQ-Upload kind=pcq pcq-rate=0 pcq-classifier=src-address

Otomatisasi dengan Ansible

• Template Jinja2 untuk konfigurasi standar per role (core/agg/access)
• Inventory dinamis dari NetBox
• Rollback otomatis jika health check gagal pasca-deploy

Hasil & Metrik

Metrik	Sebelum	Sesudah	Peningkatan
Provisioning site baru	4 jam	30 menit	87% faster
Latency antar site (rata²)	45-120ms	12-25ms	70% lower
Uptime jaringan	99.2%	99.95%	+0.75%
Bandwidth consistency	60-100%	95% ±5%	Stabil
MTTR gangguan core	2-4 jam	<15 menit	90% faster
OPEX bulanan	Baseline	-40%	Hemat signifikan

Lessons Learned

1. Invest di automation early — template + CI/CD bayar sendiri di site ke-5
2. MPLS L3VPN > VPLS untuk ISP — routing control lebih granular, scale lebih baik
3. Monitoring proactive — BGP PIC + BFD + NETCONF telemetry = deteksi <1 detik
4. Documentation as code — diagram di Git, auto-generated dari Ansible inventory
5. Training tim internal — knowledge transfer krusial, jangan hanya vendor

Referensi Teknis

• MTCINE Syllabus — BGP Advanced, MPLS L3VPN/L2VPN, EVPN, Segment Routing
• MTCTCE Syllabus — QoS HTB, PCQ, Traffic Engineering
• MTCRE Syllabus — OSPF, BGP, MPLS Fundamentals