Appendix M: Detailed Capacity Planning & Design Guide¶
9-Site Global Deployment | Hardware Selection | Performance Engineering
Sites: Mumbai | Chennai | London | Frankfurt | New Jersey | Dallas | Bangalore | Delhi | Noida
Document Information¶
| Field | Value |
|---|---|
| Document Title | Abhavtech SD-WAN Capacity Planning & Design |
| Version | 1.0 |
| Date | January 2026 |
| Organization | Abhavtech.com |
| Classification | Internal - Technical Reference |
| Status | Production Design |
Executive Summary¶
This document provides comprehensive capacity planning, hardware selection rationale, and traffic flow design for Abhavtech's 9-site Cisco Catalyst SD-WAN deployment. The design supports current requirements while accommodating 5-year growth projections from 2,500 users to 10,000 users across global operations.
Key Metrics: - Sites: 9 (6 hubs, 3 branches) - WAN Edge Devices: 15 (dual at hubs, single at branches) - Total Users: 2,500 (current) → 10,000 (Year 5) - Aggregate Bandwidth: 4,500 Mbps (current) → 30,000 Mbps (Year 5) - Software: vManage 20.15.x, vSmart 20.15.x, IOS-XE 17.15.x - Integration: SD-Access fabric handoff (independent domain model)
Table of Contents¶
- Site-Specific Analysis
- Hardware Selection Rationale
- Capacity Planning by Site
- Traffic Flow Design
- Transport Architecture
- Performance Baselines
- Growth Projections
- Design Considerations
1. Site-Specific Analysis¶
1.1 Site Overview¶
┌──────────────────────────────────────────────────────────────────────────┐
│ ABHAVTECH GLOBAL SITE TOPOLOGY │
├──────────────────────────────────────────────────────────────────────────┤
│ │
│ SD-WAN CONTROLLER PLANE │
│ ┌────────────────────────┐ │
│ │ vManage Cluster (3+3) │ │
│ │ Mumbai DC + Chennai DR │ │
│ └───────────┬────────────┘ │
│ │ │
│ ┌───────────────┼───────────────┐ │
│ │ │ │ │
│ ┌─────▼──────┐ ┌────▼──────┐ ┌────▼──────┐ │
│ │ vSmart x2 │ │ vSmart x2 │ │ vBond x2 │ │
│ │ Mumbai │ │ Chennai │ │ Cloud │ │
│ └─────┬──────┘ └────┬──────┘ └────┬──────┘ │
│ │ │ │ │
│ └──────────────┴──────────────┘ │
│ DATA PLANE │
│ │ │
│ ┌────────────────────────┼────────────────────────┐ │
│ │ │ │ │
│ ASIA-PACIFIC EMEA AMERICAS │
│ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐ │
│ │ Mumbai DC │◄─────►│ London │◄───────►│ New Jersey │ │
│ │ (Hub) │ │ (Hub) │ │ (Hub) │ │
│ │ C8500 x2 │ │ C8300 x2 │ │ C8300 x2 │ │
│ └──────┬──────┘ └──────┬──────┘ └──────┬──────┘ │
│ │ │ │ │
│ ┌──────▼──────┐ ┌──────▼──────┐ ┌──────▼──────┐ │
│ │ Chennai DR │ │ Frankfurt │ │ Dallas │ │
│ │ (Hub) │ │ (Hub) │ │ (Hub) │ │
│ │ C8500 x2 │ │ C8300 x2 │ │ C8300 x2 │ │
│ └──────┬──────┘ └─────────────┘ └─────────────┘ │
│ │ │
│ ┌──────┴────────────┐ │
│ │ │ │ │ │
│ ▼ ▼ ▼ │ │
│ ┌───┐ ┌───┐ ┌────┐ │ │
│ │BLR│ │DEL│ │NOI │ │ │
│ │C82│ │C82│ │C82L│ │ │
│ └───┘ └───┘ └────┘ │ │
│ │ │
└────────────────────────┘─────────────────────────────────────────────────┘
1.2 Detailed Site Matrix¶
| Site Code | Location | Region | Type | Users | SD-Access | WAN Edge Model | Qty | Primary Role |
|---|---|---|---|---|---|---|---|---|
| MUM | Mumbai | APAC | Data Center | 800 | Yes (Border x4) | C8500-12X4QC | 2 | Primary Hub APAC |
| CHN | Chennai | APAC | Data Center | 600 | Yes (Border x4) | C8500-12X4QC | 2 | DR + Secondary Hub |
| LON | London | EMEA | Regional HQ | 150 | No | C8300-2N2S-6T | 2 | Primary Hub EMEA |
| FRA | Frankfurt | EMEA | Regional Office | 50 | No | C8300-2N2S-6T | 2 | Secondary Hub EMEA |
| NJ | New Jersey | Americas | Regional HQ | 30 | No | C8300-2N2S-6T | 2 | Primary Hub Americas |
| DAL | Dallas | Americas | Regional Office | 20 | No | C8300-2N2S-6T | 2 | Secondary Hub Americas |
| BLR | Bangalore | APAC | Branch | 400 | No | C8200-1N-4T | 1 | Large Branch |
| DEL | Delhi | APAC | Branch | 350 | No | C8200-1N-4T | 1 | Large Branch |
| NOI | Noida | APAC | Branch | 100 | No | C8200L-1N-4T | 1 | Medium Branch |
2. Hardware Selection Rationale¶
2.1 Controller Infrastructure¶
vManage Cluster Sizing¶
Selected Configuration: Large Scale
| Component | Specification | Rationale |
|---|---|---|
| Nodes | 3 (active cluster) + 3 (DR) | HA + disaster recovery |
| vCPU | 32 per node | Support 501-2000 edges |
| RAM | 128 GB per node | Analytics + logging |
| Storage | 2 TB SSD per node | Statistics database |
| Deployment | ESXi hypervisor | Existing infrastructure |
| Location | Mumbai (Primary), Chennai (DR) | Regional distribution |
Growth Headroom: - Current: 15 WAN edges - Year 5 projection: 85 WAN edges - Controller capacity: Up to 2,000 edges - Headroom: 95% ✓
vSmart Controller Sizing¶
Deployment Model: Regional Distribution
| Location | Quantity | vCPU | RAM | Edges Supported | OMP Routes |
|---|---|---|---|---|---|
| Mumbai DC | 2 | 16 | 32 GB | 500-2000 | 250,000 |
| Chennai DR | 2 | 16 | 32 GB | 500-2000 | 250,000 |
| Total | 4 | 64 | 128 GB | 2000-4000 | 500K |
Rationale: - 4 controllers provide 2:1 redundancy - Each WAN edge connects to 2 vSmarts minimum - Regional placement reduces OMP latency - Mumbai controllers serve APAC/EMEA - Chennai controllers serve Americas + backup
vBond Validator¶
Deployment Model: Cloud-Hosted
| Attribute | Specification |
|---|---|
| Quantity | 2 (active-active) |
| vCPU | 2 per instance |
| RAM | 4 GB per instance |
| Location | AWS (Mumbai region), AWS (Singapore region) |
| Public IP | Yes (required for ZTP) |
| DNS | vbond.abhavtech.com (Round-robin) |
2.2 WAN Edge Selection by Site¶
Mumbai DC (MUM) - Primary Data Center¶
Selected Platform: Cisco Catalyst 8500-12X4QC (x2)
| Requirement | Specification | C8500 Capability | Match |
|---|---|---|---|
| Throughput | 5,000 Mbps (current), 16,000 Mbps (Y5) | 100 Gbps | ✓ |
| Users | 800 (current), 2,000 (Y5) | 10,000+ | ✓ |
| Tunnels | 200 (current), 1,000 (Y5) | 16,000 | ✓ |
| Sessions | 500K (current), 2M (Y5) | 2M | ✓ |
| High Availability | Active-active dual router | Supported | ✓ |
| SD-Access Handoff | 4x Border nodes | 12x 10GbE ports | ✓ |
| Encryption | AES-256-GCM hardware | Hardware crypto | ✓ |
| UTD | Snort 3.0 IPS | Full UTD support | ✓ |
| 5G Backup | Yes | Cellular module support | ✓ |
Interface Utilization:
Mumbai WAN Edge 1 (MUM-WAN-01):
┌────────────────────────────────────────────────────────────┐
│ C8500-12X4QC Interface Layout │
├────────────────────────────────────────────────────────────┤
│ │
│ Te0/0/0 [10G] ──► MPLS Transport (Tata, 2 Gbps) │
│ Te0/0/1 [10G] ──► DIA Transport (Tata, 2 Gbps) │
│ Cell0/4/0 ─────► 5G Backup (Jio, 1 Gbps) │
│ │
│ Te0/1/0 [10G] ──► SD-Access Border-1 (VRF handoff) │
│ Te0/1/1 [10G] ──► SD-Access Border-2 (VRF handoff) │
│ Te0/1/2 [10G] ──► SD-Access Border-3 (VRF handoff) │
│ Te0/1/3 [10G] ──► SD-Access Border-4 (VRF handoff) │
│ │
│ Te0/2/0 [10G] ──► Campus Core (Management) │
│ │
│ Fo0/3/0 [40G] ──► Reserved for future expansion │
│ │
└────────────────────────────────────────────────────────────┘
Port Utilization: 9 of 16 ports (56%)
Chennai DR (CHN) - Disaster Recovery + Secondary Hub¶
Selected Platform: Cisco Catalyst 8500-12X4QC (x2)
| Requirement | Specification | Match |
|---|---|---|
| Throughput | 4,000 Mbps (current), 12,000 Mbps (Y5) | ✓ (same as Mumbai) |
| HA | Active-active | ✓ |
| SD-Access | 4x Border nodes | ✓ |
| Role | DR + active hub for branches | ✓ |
Note: Chennai uses identical hardware to Mumbai for: - Simplified spare parts inventory - DR failover capability - Consistent performance profile
London (LON) - Primary Hub EMEA¶
Selected Platform: Cisco Catalyst 8300-2N2S-6T (x2)
| Requirement | Specification | C8300 Capability | Match |
|---|---|---|---|
| Throughput | 1,000 Mbps (current), 3,000 Mbps (Y5) | 10 Gbps | ✓ |
| Users | 150 (current), 400 (Y5) | 2,000+ | ✓ |
| Tunnels | 50 (current), 200 (Y5) | 8,000 | ✓ |
| HA | Active-active | Supported | ✓ |
| Ports | 3 transports + 2 LAN | 2xNIM + 6x1G | ✓ |
Interface Layout:
London WAN Edge 1 (LON-WAN-01):
┌────────────────────────────────────────────────────────────┐
│ C8300-2N2S-6T Interface Layout │
├────────────────────────────────────────────────────────────┤
│ │
│ NIM Slot 0: NIM-2GE-CU-SFP │
│ ├─ Gi0/0/0 ──► MPLS Transport (BT, 500 Mbps) │
│ └─ Gi0/0/1 ──► DIA Transport (BT, 500 Mbps) │
│ │
│ NIM Slot 1: NIM-LTEA-EA (LTE Americas/EMEA) │
│ └─ Cell0/1/0 ──► 5G Backup (EE, 500 Mbps) │
│ │
│ Onboard Ports: │
│ ├─ Gi0/0/2 ──► Campus Core (VLAN trunk) │
│ ├─ Gi0/0/3 ──► Management VLAN │
│ └─ Gi0/0/4-5 ──► Reserved │
│ │
└────────────────────────────────────────────────────────────┘
Frankfurt, New Jersey, Dallas (FRA, NJ, DAL) - Regional Hubs¶
Selected Platform: Cisco Catalyst 8300-2N2S-6T (x2 each site)
Rationale: Same as London - Consistent platform across all regional hubs - Simplified operations and training - Common spare parts pool - Right-sized for 50-500 user sites
Bangalore, Delhi (BLR, DEL) - Large Branches¶
Selected Platform: Cisco Catalyst 8200-1N-4T (x1 each)
| Requirement | Specification | C8200 Capability | Match |
|---|---|---|---|
| Throughput | 300-650 Mbps (current-Y3) | 1-2 Gbps | ✓ |
| Users | 350-400 | 1,000+ | ✓ |
| HA | Single router + transport redundancy | Supported | ✓ |
| Cost | Budget-conscious | Mid-range | ✓ |
Interface Layout:
Bangalore WAN Edge (BLR-WAN-01):
┌────────────────────────────────────────────────────────────┐
│ C8200-1N-4T Interface Layout │
├────────────────────────────────────────────────────────────┤
│ │
│ NIM Slot 0: NIM-2GE-CU-SFP + NIM-LTEA-LA │
│ ├─ Gi0/0/0 ──► DIA Transport (Jio, 200 Mbps) │
│ └─ Cell0/0/1 ──► LTE Backup (Airtel, 100 Mbps) │
│ │
│ Onboard Ports: │
│ ├─ Gi0/0/1 ──► Campus Distribution Switch │
│ ├─ Gi0/0/2 ──► Management VLAN │
│ └─ Gi0/0/3-4 ──► Reserved │
│ │
└────────────────────────────────────────────────────────────┘
Why No Dual Routers: - Cost optimization for branch sites - Transport redundancy (DIA + LTE) provides adequate resilience - MTTR target: 4 hours (spare on-site) - Business criticality: Medium (can tolerate brief outages)
Noida (NOI) - Medium Branch¶
Selected Platform: Cisco Catalyst 8200L-1N-4T (x1)
| Requirement | Specification | C8200L Capability | Match |
|---|---|---|---|
| Throughput | 100-500 Mbps | 500 Mbps - 1 Gbps | ✓ |
| Users | 100 (current), 250 (Y5) | 500+ | ✓ |
| Cost | Most budget-friendly | Entry-level | ✓ |
| Complexity | Simplified operations | Limited features OK | ✓ |
3. Capacity Planning by Site¶
3.1 Mumbai DC (MUM) - Detailed Capacity Analysis¶
Current State (Year 0 - 2026)¶
User Profile: - Total users: 800 - Knowledge workers: 600 (75%) - Power users (engineering): 150 (19%) - Executives: 50 (6%)
Bandwidth Calculation:
| Component | Quantity | BW per Unit | Subtotal |
|---|---|---|---|
| User Traffic | |||
| Knowledge workers | 600 | 2 Mbps | 1,200 Mbps |
| Power users | 150 | 4 Mbps | 600 Mbps |
| Executives | 50 | 3 Mbps | 150 Mbps |
| User Subtotal | 1,950 Mbps | ||
| Application Traffic | |||
| Voice (10% concurrent) | 80 calls | 100 Kbps | 8 Mbps |
| Video (5% concurrent) | 40 sessions | 3 Mbps | 120 Mbps |
| SAP ERP | 200 concurrent | 100 Kbps | 20 Mbps |
| Citrix VDI | 100 sessions | 150 Kbps | 15 Mbps |
| File sync (OneDrive) | 400 users | 500 Kbps | 200 Mbps |
| App Subtotal | 363 Mbps | ||
| Inter-Site Traffic | |||
| To Chennai DR | 800 Mbps | ||
| To other hubs | 500 Mbps | ||
| From branches | 1,500 Mbps | ||
| Inter-Site Subtotal | 2,800 Mbps | ||
| Cloud Traffic | |||
| M365 (via DIA) | 500 Mbps | ||
| AWS (Mumbai region) | 300 Mbps | ||
| Cloud Subtotal | 800 Mbps | ||
| Total Before Overhead | 5,913 Mbps | ||
| 20% Overhead | 1,183 Mbps | ||
| TOTAL CALCULATED | 7,096 Mbps |
Provisioned Transport:
| Transport | Bandwidth | Utilization @ Peak | Headroom |
|---|---|---|---|
| MPLS | 2,000 Mbps | 70% (1,400 Mbps) | 30% |
| DIA | 2,000 Mbps | 60% (1,200 Mbps) | 40% |
| 5G | 1,000 Mbps | 10% (100 Mbps, backup) | 90% |
| Total | 5,000 Mbps | 54% | 46% |
Platform Validation: - C8500-12X4QC rated: 100 Gbps - Required (with IPsec + UTD): ~7 Gbps effective - C8500 effective (IPsec + UTD): ~55 Gbps - Headroom: 87% ✓
Year 3 Projection (2029)¶
Users: 800 → 2,000 (150% growth) Calculated Bandwidth: 7,096 Mbps × 1.5 ×1.2 (traffic growth) = 12,773 Mbps
Provisioned Transport (Year 3):
| Transport | Bandwidth | Cost/Month |
|---|---|---|
| MPLS (retained) | 2,000 Mbps | $XX,XXX |
| DIA (upgraded) | 5,000 Mbps | $XX,XXX |
| 5G (upgraded) | 2,000 Mbps | $X,XXX |
| Total | 9,000 Mbps | $XX,XXX |
Platform: C8500 still adequate (55 Gbps effective >> 12.7 Gbps required)
Year 5 Projection (2031)¶
Users: 800 → 3,000 Calculated Bandwidth: ~18,000 Mbps
Recommendation: - Option A: Upgrade to dual 100G interfaces - Option B: Add third C8500 (3-way active-active) - Selected: Option A (simpler, lower OPEX)
3.2 Chennai DR (CHN)¶
Current State¶
| Metric | Value | Notes |
|---|---|---|
| Users | 600 | 75% of Mumbai |
| Calculated BW | 5,327 Mbps | Similar profile to Mumbai |
| Provisioned Transport | MPLS 2G + DIA 1.5G + 5G 500M | 4,000 Mbps total |
| Peak Utilization | 62% | |
| Platform | C8500-12X4QC x2 | Same as Mumbai (DR parity) |
Special Considerations¶
Disaster Recovery Role: - Must handle 100% Mumbai traffic if DR invoked - Provisioned for 80% Mumbai + 100% local = 9,000 Mbps minimum - Current: 4,000 Mbps (will increase to 9G by Y2)
Replication Traffic: - Mumbai ↔ Chennai: 800 Mbps dedicated - Database replication: Real-time - Backup replication: Off-peak
3.3 Regional Hubs (LON, FRA, NJ, DAL)¶
London (LON)¶
| Metric | Current | Year 3 | Year 5 |
|---|---|---|---|
| Users | 150 | 400 | 800 |
| User BW | 435 Mbps | 1,160 Mbps | 2,320 Mbps |
| App BW | 109 Mbps | 290 Mbps | 580 Mbps |
| Cloud BW | 200 Mbps | 500 Mbps | 1,000 Mbps |
| Inter-Hub | 100 Mbps | 300 Mbps | 600 Mbps |
| Total Calculated | 844 Mbps | 2,250 Mbps | 4,500 Mbps |
| Provisioned | 1,000 Mbps | 3,000 Mbps | 5,000 Mbps |
| Platform | C8300 (10G capable) | C8300 | C8300 or C8500 |
Frankfurt (FRA), New Jersey (NJ), Dallas (DAL)¶
Similar sizing methodology applied:
| Site | Users (Y0/Y5) | BW (Y0/Y5) | Platform |
|---|---|---|---|
| Frankfurt | 50 / 400 | 500M / 3G | C8300 / C8300 |
| New Jersey | 30 / 500 | 300M / 4G | C8300 / C8300 |
| Dallas | 20 / 300 | 200M / 2.5G | C8300 / C8300 |
3.4 Branches (BLR, DEL, NOI)¶
Bangalore (BLR)¶
Current State:
| Component | Calculation | Result |
|---|---|---|
| Users | 400 | |
| User traffic | 400 × 2 Mbps | 800 Mbps |
| Application traffic | Voice: 3M, Video: 20M, Apps: 25M | 48 Mbps |
| Cloud traffic | M365: 100M | 100 Mbps |
| Total (with overhead) | (800+48+100) × 1.2 | 1,138 Mbps |
| Provisioned | DIA 200M + LTE 100M | 300 Mbps |
| Utilization @ Peak | ~75% |
Analysis: - Current provisioning appears undersized - Recommendation: Upgrade DIA to 500 Mbps by Year 1 - Platform: C8200-1N-4T (2 Gbps capable) adequate through Year 5
Delhi (DEL)¶
| Metric | Current | Year 3 |
|---|---|---|
| Users | 350 | 700 |
| Calculated BW | 1,000 Mbps | 2,000 Mbps |
| Provisioned | 300 Mbps | 1,000 Mbps |
| Platform | C8200 (adequate) | C8200 (adequate) |
Noida (NOI)¶
| Metric | Current | Year 5 |
|---|---|---|
| Users | 100 | 250 |
| Calculated BW | 286 Mbps | 715 Mbps |
| Provisioned | 200 Mbps | 500 Mbps |
| Platform | C8200L (500M-1G) | C8200 (upgrade needed Y4) |
4. Traffic Flow Design¶
4.1 Hub-to-Hub Traffic Flows¶
Inter-Hub Connectivity Matrix¶
HUB-TO-HUB TRAFFIC MATRIX
┌──────────────────────────────────────────────────────────────────────────┐
│ │
│ MUM ─────── CHN │
│ │ \ / │ │
│ 800M│ \300M / │600M │
│ │ \ / │ │
│ │ \ / │ │
│ │ X │ │
│ │ / \ │ │
│ │ / \ │ │
│ │ / \ │ │
│ │ / \ │ │
│ LON ─────── FRA │
│ │ 200M │ │
│ │ │ │
│ 150M │ │ 100M │
│ │ │ │
│ NJ ────────DAL │
│ 150M │
│ │
│ TRAFFIC VOLUMES (Mbps, Current): │
│ ──────────────────────────────────── │
│ Mumbai ↔ Chennai: 800 (DB replication, shared services) │
│ Mumbai ↔ London: 300 (APAC-EMEA integration) │
│ Mumbai ↔ NJ: 200 (APAC-Americas integration) │
│ Chennai ↔ London: 200 (Backup path) │
│ Chennai ↔ NJ: 150 (Backup path) │
│ London ↔ Frankfurt: 200 (Regional collaboration) │
│ London ↔ NJ: 100 (Trans-Atlantic) │
│ NJ ↔ Dallas: 150 (Americas regional) │
│ │
└──────────────────────────────────────────────────────────────────────────┘
Hub-to-Hub Traffic Characteristics¶
| Flow | Volume | Type | Transport Preference | SLA |
|---|---|---|---|---|
| MUM ↔ CHN | 800M | DB replication, shared services | MPLS (primary) | <10ms, <0.1% loss |
| MUM ↔ LON | 300M | Integration, file sharing | MPLS preferred | <120ms |
| MUM ↔ NJ | 200M | Integration, reporting | MPLS preferred | <180ms |
| LON ↔ FRA | 200M | EMEA regional | MPLS/DIA | <20ms |
| NJ ↔ DAL | 150M | Americas regional | MPLS/DIA | <30ms |
4.2 Branch-to-Hub Traffic Flows¶
Branch Traffic Distribution¶
BRANCH-TO-HUB TRAFFIC PATTERNS
┌──────────────────────────────────────────────────────────────────────────┐
│ │
│ Bangalore (BLR) ───► Mumbai (Primary): 200 Mbps │
│ └──► Chennai (Backup): 50 Mbps │
│ │
│ Delhi (DEL) ───────► Mumbai (Primary): 180 Mbps │
│ └──► Chennai (Backup): 45 Mbps │
│ │
│ Noida (NOI) ───────► Mumbai (Primary): 80 Mbps │
│ └──► Chennai (Backup): 20 Mbps │
│ │
│ TRAFFIC BREAKDOWN BY DESTINATION: │
│ │
│ ┌───────────────────────────────────────────────────────┐ │
│ │ Bangalore (400 users, 300 Mbps provisioned): │ │
│ ├───────────────────────────────────────────────────────┤ │
│ │ → Mumbai DC (SAP, file servers): 200 Mbps │ │
│ │ → Chennai DR (backup): 50 Mbps │ │
│ │ → Internet (M365, cloud): 100 Mbps (local breakout) │ │
│ │ → Overhead/Other: 50 Mbps │ │
│ └───────────────────────────────────────────────────────┘ │
│ │
└──────────────────────────────────────────────────────────────────────────┘
Branch Application Profiles¶
| Application | Destination | % of BW | Transport | QoS Class |
|---|---|---|---|---|
| SAP ERP | Mumbai DC | 30% | MPLS preferred | Business-Critical |
| File Servers | Mumbai DC | 25% | MPLS/DIA | Default |
| M365 | Internet | 20% | DIA (local breakout) | Default |
| Voice | Mumbai UC | 5% | Best path (AAR) | Real-Time |
| Video Conf | Cloud/Internet | 10% | DIA preferred | Real-Time |
| Backup | Chennai DR | 5% | Best-effort | Best-Effort |
| Other | Mixed | 5% | Default | Default |
4.3 Application-Aware Routing (AAR) Configuration¶
SLA Classes¶
┌──────────────────────────────────────────────────────────────────────────┐
│ ABHAVTECH APPLICATION SLA CLASSES │
├──────────────────────────────────────────────────────────────────────────┤
│ │
│ REAL-TIME (Voice, Video): │
│ ──────────────────────────── │
│ • Latency: <150ms │
│ • Jitter: <30ms │
│ • Loss: <1% │
│ • Applications: Webex, Teams, SIP │
│ • Transport: Any (automatically selects best) │
│ • Failover: Immediate (<5 seconds) │
│ │
│ BUSINESS-CRITICAL (SAP, Oracle, Citrix): │
│ ───────────────────────────────────────── │
│ • Latency: <200ms │
│ • Jitter: <50ms │
│ • Loss: <2% │
│ • Applications: SAP, Oracle ERP, Citrix VDI │
│ • Transport: Prefer MPLS, allow DIA │
│ • Failover: Switch within 60 seconds │
│ │
│ DEFAULT (General business): │
│ ─────────────────────────── │
│ • Latency: <300ms │
│ • Jitter: <100ms │
│ • Loss: <5% │
│ • Applications: HTTP, HTTPS, email │
│ • Transport: Load balance across all │
│ • Failover: Switch within 120 seconds │
│ │
│ BEST-EFFORT (Bulk data): │
│ ─────────────────────────── │
│ • Latency: No limit │
│ • Jitter: No limit │
│ • Loss: <10% │
│ • Applications: Backup, file sync (non-interactive) │
│ • Transport: Least-cost path │
│ • Failover: None (wait for recovery) │
│ │
└──────────────────────────────────────────────────────────────────────────┘
Mumbai to London - AAR Example¶
Scenario: Multiple applications from MUM to LON
| Application | SLA Class | Normal Path | Fallback Path | Path Logic |
|---|---|---|---|---|
| Webex | Real-Time | MPLS (8ms jitter) | DIA (12ms jitter) | Latency OK on both, prefer MPLS for stability |
| SAP | Bus-Critical | MPLS (10ms RTT) | DIA (15ms RTT) | MPLS meets <200ms easily |
| File Sync | Default | DIA (higher BW) | MPLS | DIA has more available bandwidth |
| Backup | Best-Effort | DIA (off-peak) | - | Use cheapest path |
Failover Example:
Normal State:
Webex ─────MPLS─────► London (Latency: 125ms, Loss: 0.1%)
Event: MPLS latency spikes to 200ms (SLA violation)
AAR Action (within 5 seconds):
Webex ─────DIA──────► London (Latency: 140ms, Loss: 0.3%)
Recovery: After MPLS stabilizes for 10 minutes, traffic returns
5. Transport Architecture¶
5.1 Transport Strategy by Site¶
Hub Sites (MUM, CHN, LON, FRA, NJ, DAL)¶
Triple Transport Model:
| Transport | Bandwidth | Monthly Cost | Purpose | Weight |
|---|---|---|---|---|
| MPLS | 500M-2G | $XXK-25K | Primary (critical apps) | 1 |
| DIA | 500M-2G | $XXX-X,XXX | Primary (cloud, Internet) | 1 |
| 5G/LTE | 100M-1G | $XXX-X,XXX | Backup (last-resort) | 50 |
Mumbai Transport Details:
| Circuit ID | Provider | Type | Bandwidth | IP Subnet | Cost/Month | SLA |
|---|---|---|---|---|---|---|
| MPLS-MUM-001 | Tata Communications | MPLS L3VPN | 2 Gbps | 198.51.100.0/30 | $XX,XXX | 99.9% |
| DIA-MUM-001 | Tata Communications | Dedicated Internet | 2 Gbps | 203.0.113.0/30 | $X,XXX | 99.5% |
| 5G-MUM-001 | Jio | 5G SA | 1 Gbps (burst) | DHCP | $X,XXX | 99% |
Branch Sites (BLR, DEL, NOI)¶
Dual Transport Model:
| Site | Primary | Bandwidth | Secondary | Bandwidth | Total Cost |
|---|---|---|---|---|---|
| Bangalore | Jio DIA | 200M | Airtel LTE | 100M | $XXX/mo |
| Delhi | Airtel DIA | 200M | Jio LTE | 100M | $XXX/mo |
| Noida | Tata DIA | 100M | Jio LTE | 50M | $XXX/mo |
Cost Comparison vs. MPLS:
| Site | Old MPLS Cost | New SD-WAN Cost | Savings | % Reduction |
|---|---|---|---|---|
| Bangalore | $X,XXX/mo | $XXX/mo | $X,XXX/mo | 80% |
| Delhi | $X,XXX/mo | $XXX/mo | $X,XXX/mo | 74% |
| Noida | $X,XXX/mo | $XXX/mo | $X,XXX/mo | 73% |
5.2 Transport Performance Baselines¶
Expected SLA by Transport Type¶
| Transport | Site | Latency (to MUM) | Jitter | Loss | Notes |
|---|---|---|---|---|---|
| MPLS | |||||
| Mumbai ↔ Chennai | MPLS | 5ms | 1ms | 0.01% | Intra-region, excellent |
| Mumbai ↔ Bangalore | MPLS | 12ms | 2ms | 0.05% | Regional fiber |
| Mumbai ↔ London | MPLS | 120ms | 5ms | 0.1% | Trans-continental |
| DIA | |||||
| Mumbai ↔ Chennai | DIA | 8ms | 2ms | 0.02% | Good peering |
| Mumbai ↔ Bangalore | DIA | 20ms | 8ms | 0.2% | Acceptable |
| Mumbai ↔ London | DIA | 140ms | 15ms | 0.3% | Variable routing |
| 5G/LTE | |||||
| Mumbai local | 5G | 25ms | 10ms | 0.5% | Backup only |
| Bangalore | LTE | 35ms (to MUM) | 15ms | 1% | Last-resort |
6. Performance Baselines¶
6.1 Application Performance Targets¶
By Application Type¶
| Application | Latency Target | Throughput | Concurrent Users | Platform Requirement |
|---|---|---|---|---|
| SAP ERP | <100ms | 50-100 Kbps/user | 200 @ Mumbai | Low latency path |
| M365 | <50ms (to cloud) | Variable | 800 @ Mumbai | DIA local breakout |
| Webex | <150ms | 2-4 Mbps | 40 concurrent | Real-time QoS |
| Citrix VDI | <50ms | 150 Kbps/session | 100 concurrent | Consistent latency |
| File Sharing | <200ms | Bursty | 400 concurrent | High bandwidth |
6.2 Feature Impact on Performance¶
C8500 (Mumbai/Chennai)¶
| Feature Enabled | Raw Throughput | Effective Throughput | Impact |
|---|---|---|---|
| Baseline | 100 Gbps | 100 Gbps | 0% |
| + IPsec AES-256 | 100 Gbps | 85 Gbps | -15% |
| + UTD (Snort 3.0) | 85 Gbps | 55 Gbps | -35% |
| Final | 100 Gbps | ~55 Gbps | -45% |
Validation: - Mumbai requirement: 7 Gbps (current), 18 Gbps (Year 5) - C8500 effective: 55 Gbps - Headroom: 67% (Year 5) ✓
C8300 (Regional Hubs)¶
| Feature Enabled | Raw Throughput | Effective Throughput |
|---|---|---|
| Baseline | 10 Gbps | 10 Gbps |
| + IPsec + UTD | 10 Gbps | 5.5 Gbps |
Validation: - London requirement: 2.25 Gbps (Year 3) - C8300 effective: 5.5 Gbps - Headroom: 59% ✓
C8200 (Branches)¶
| Feature Enabled | Raw Throughput | Effective Throughput |
|---|---|---|
| Baseline | 1-2 Gbps | 1-2 Gbps |
| + IPsec + UTD | 1-2 Gbps | 1.1 Gbps |
Validation: - Bangalore requirement: 1.1 Gbps - C8200 effective: 1.1 Gbps - Headroom: Minimal (plan upgrade Year 3) ⚠️
7. Growth Projections¶
7.1 5-Year Capacity Roadmap¶
Aggregate Deployment Statistics¶
| Metric | Y0 (2026) | Y1 (2027) | Y3 (2029) | Y5 (2031) | CAGR |
|---|---|---|---|---|---|
| Sites | 9 | 12 | 25 | 50 | 40% |
| WAN Edges | 15 | 22 | 45 | 85 | 42% |
| Users | 2,500 | 3,500 | 6,000 | 10,000 | 32% |
| Bandwidth (Mbps) | 4,500 | 7,000 | 15,000 | 30,000 | 46% |
| VPNs/Segments | 5 | 8 | 12 | 15 | 25% |
| Policies | 50 | 100 | 200 | 350 | 48% |
When to Upgrade¶
| Site | Component | Current | Upgrade Trigger | Upgrade To | Timeline |
|---|---|---|---|---|---|
| Mumbai | Transport | 5G total | >70% utilization | 10G MPLS + 5G DIA | Year 3 |
| Chennai | Transport | 4G total | DR capacity | Match Mumbai | Year 2 |
| London | Platform | C8300 | >60% throughput | C8500 or 2nd C8300 | Year 4 |
| Bangalore | Platform | C8200 | >80% throughput | C8300 | Year 3 |
| Noida | Platform | C8200L | >500M sustained | C8200 | Year 4 |
| Controllers | vManage | 32v/128G | >500 edges | No change needed | N/A |
| Controllers | vSmart | 16v/32G | >2000 edges | No change needed | N/A |
7.2 Cost Projections¶
Capital Expenditure (CapEx)¶
| Year | Hardware | Licenses | Services | Total CapEx | Notes |
|---|---|---|---|---|---|
| Y0 | $XXXK | $XXXK | $XXXK | $XXXK | Initial deployment |
| Y1 | $XXXK | $XXK | $XXK | $XXXK | 3 new sites |
| Y2 | $XXK | $XXK | $XXK | $XXXK | Chennai upgrade |
| Y3 | $XXXK | $XXXK | $XXK | $XXXK | 13 new sites |
| Y5 | $XXXK | $XXXK | $XXK | $XXXK | 25 new sites |
| 5-Year Total | $X.XXM | $XXXK | $XXXK | $X.XXM |
Operational Expenditure (OpEx)¶
| Category | Y0 | Y3 | Y5 | Notes |
|---|---|---|---|---|
| Transport Costs | $XXXK | $XXXK | $X.XM | DIA scales with BW |
| DNA Licensing | $XXXK | $XXXK | $XXXK | Per-device + bandwidth |
| Support/Maintenance | $XXXK | $XXXK | $XXXK | 20% of CapEx |
| Operational Labor | $XXXK | $XXXK | $XXXK | NOC + engineering |
| Annual OpEx | $XXXK | $X.XM | $X.XM |
vs. MPLS Baseline: - Y0 MPLS: $X.XM/year - Y0 SD-WAN: $XXXK/year - Savings: 61% ($X.XXM/year)
8. Design Considerations¶
8.1 Key Design Decisions¶
Decision 1: Dual WAN Edge at Hubs Only¶
Rationale: - Hubs aggregate traffic from multiple branches - Hub failures impact many users (blast radius) - Cost justified by criticality - Branches have transport redundancy (DIA + LTE) for resiliency
Trade-off: - Branch MTTR: 4 hours (acceptable for business) - Spare hardware staged at regional hubs
Decision 2: MPLS Retention Strategy¶
Kept: - Mumbai ↔ Chennai (critical DB replication) - Hub-to-hub primary paths - Cost: $XXXK/year
Migrated: - All branch MPLS circuits - Branch-to-hub uses DIA + LTE - Savings: $XXXK/year
Rationale: - Hub-to-hub MPLS provides consistent low latency - Cost reduction while maintaining critical performance
Decision 3: DIA Local Breakout for SaaS¶
Decision: - M365, Webex, cloud apps bypass hub - Direct from branch to Internet - Umbrella DNS filtering at all sites
Benefits: - Reduced hub backhaul (saves ~500 Mbps @ Mumbai) - Better user experience (lower latency to cloud) - Cost-effective
Decision 4: No Branch-to-Branch Direct Tunnels¶
Decision: - All branch traffic routes via hubs - No on-demand branch-to-branch tunnels (at launch)
Rationale: - Minimal branch-to-branch traffic (estimated <5%) - Simplified troubleshooting - Can enable on-demand later if needed
Decision 5: Hardware Crypto Mandatory¶
Decision: - All platforms must have hardware crypto acceleration - AES-256-GCM encryption enabled by default
Rationale: - Minimal performance impact with HW crypto - Compliance requirements (data-in-transit encryption) - No CPU penalty
8.2 SD-Access Integration Considerations¶
Mumbai/Chennai SD-Access Handoff¶
Architecture: - Independent domain model (SD-Access ↔ SD-WAN are separate) - L3 VRF handoff via eBGP - SD-Access Border nodes: 4 per site (C9500-40X) - Connectivity: 4x 10GbE links per WAN Edge
VN-to-VPN Mapping:
| SD-Access VN | VN ID | SD-WAN VPN | VPN ID | Purpose |
|---|---|---|---|---|
| Corporate | 4097 | Corporate | 10 | Employee endpoints |
| Guest | 4098 | Guest | 20 | Guest WiFi |
| Servers | 4099 | Servers | 30 | DC servers |
| Voice | 4100 | Voice | 40 | IP phones, UC |
| IoT | 4101 | IoT | 50 | Smart building |
Route Exchange: - SD-Access exports: Campus subnets (LISP EIDs) - SD-WAN exports: WAN site subnets - Filtering: Deny default routes, summarize where possible
SGT Propagation: - Inline tagging across WAN - TrustSec enabled on WAN Edge - SGT preserved end-to-end
8.3 High Availability Considerations¶
Failure Scenarios and Recovery¶
| Scenario | Detection Time | Failover Time | Total RTO | Impact |
|---|---|---|---|---|
| Transport link failure | 3s (BFD) | <1s | <5s | Seamless |
| WAN Edge failure (hub) | 3s (BFD) | <2s | <10s | Brief disruption |
| vSmart failure | 30s (OMP keepalive) | <30s | <60s | No data plane impact |
| vManage failure | N/A | N/A | 0s | No impact (data plane) |
| Site-level failure (hub) | Variable | <30s | <2 min | Route to backup hub |
BFD Timer Settings¶
| Link Type | Hello Interval | Multiplier | Failure Detection |
|---|---|---|---|
| MPLS (stable) | 1000ms | 3 | 3 seconds |
| DIA (variable) | 1000ms | 7 | 7 seconds |
| LTE (high variance) | 1000ms | 10 | 10 seconds |
Rationale: - Tighter timers for stable links (MPLS) - Looser timers for variable links (LTE) to avoid false positives
8.4 Security Considerations¶
Encryption¶
Control Plane: - DTLS 1.2 (all control connections) - Certificate-based authentication (PKI) - Perfect Forward Secrecy (PFS)
Data Plane: - IPsec ESP (all data traffic) - AES-256-GCM (authenticated encryption) - Hardware offload (no CPU penalty) - Anti-replay protection (64-bit sequence)
Segmentation¶
VPN Isolation: - VPN 10 (Corporate) isolated from VPN 20 (Guest) - Inter-VPN routing disabled by default - Firewall inspection for any inter-VPN traffic
Threat Prevention¶
UTD (Unified Threat Defense): - Snort 3.0 IPS at hub sites - Signature-based + behavioral detection - Auto-update (daily) - Performance: ~35-40% throughput impact (acceptable)
URL Filtering: - Cisco Umbrella at all sites - DNS-layer security - SaaS security for M365, Salesforce
Appendix A: Hardware BOMs¶
Mumbai DC (MUM) - Bill of Materials¶
| Part Number | Description | Quantity | Unit Price | Extended |
|---|---|---|---|---|
| C8500-12X4QC | Catalyst 8500 chassis | 2 | $XX,XXX | $XXX,XXX |
| PWR-8500-AC | 8500 AC power supply | 4 | $X,XXX | $X,XXX |
| NIM-LTEA-EA | 5G LTE module | 2 | $X,XXX | $X,XXX |
| SFP-10G-SR | 10G SFP+ transceiver | 20 | $XXX | $XX,XXX |
| CON-SNT-C85012X | SmartNet 8x5xNBD | 2 | $X,XXX | $XX,XXX |
| Subtotal Mumbai | $XXX,XXX |
Bangalore (BLR) - Bill of Materials¶
| Part Number | Description | Quantity | Unit Price | Extended |
|---|---|---|---|---|
| C8200-1N-4T | Catalyst 8200 chassis | 1 | $X,XXX | $X,XXX |
| NIM-2GE-CU-SFP | Dual GE NIM | 1 | $XXX | $XXX |
| NIM-LTEA-LA | 4G LTE module | 1 | $X,XXX | $X,XXX |
| CON-SNT-C82001N | SmartNet 8x5xNBD | 1 | $XXX | $XXX |
| Subtotal Bangalore | $XX,XXX |
Total Deployment Hardware Cost¶
| Site Tier | Sites | Cost per Site | Total |
|---|---|---|---|
| Tier 1 (C8500 x2) | 2 | $XXXK | $XXXK |
| Tier 2 (C8300 x2) | 4 | $XXK | $XXXK |
| Tier 3 (C8200 x1) | 2 | $XXK | $XXK |
| Tier 4 (C8200L x1) | 1 | $XK | $XK |
| Controllers (vManage, vSmart, vBond) | - | $XXK | $XXK |
| Grand Total | 9 | $XXXK |
Appendix B: Interface Addressing Plan¶
Mumbai WAN Edge (MUM-WAN-01)¶
| Interface | Description | IP Address | Subnet Mask | VRF/VPN | VLAN | Notes |
|---|---|---|---|---|---|---|
| Te0/0/0 | MPLS Transport | 198.51.100.1 | /30 | VPN 0 | - | Tata MPLS |
| Te0/0/1 | DIA Transport | 203.0.113.1 | /30 | VPN 0 | - | Tata DIA |
| Cell0/4/0 | 5G Backup | DHCP | - | VPN 0 | - | Jio 5G |
| Te0/1/0 | SD-Access Border-1 | 10.252.10.1 | /30 | VPN 10 | - | Corporate |
| Te0/1/1 | SD-Access Border-2 | 10.252.10.5 | /30 | VPN 20 | - | Guest |
| Te0/½ | SD-Access Border-3 | 10.252.10.9 | /30 | VPN 30 | - | Servers |
| Te0/⅓ | SD-Access Border-4 | 10.252.10.13 | /30 | VPN 40 | - | Voice |
| Loopback0 | System IP | 10.100.1.1 | /32 | - | - | OMP/BFD |
System IP and Site ID Allocation¶
| Site | Site ID | System IP (Edge 1) | System IP (Edge 2) | Region |
|---|---|---|---|---|
| Mumbai | 100 | 10.100.1.1 | 10.100.1.2 | APAC |
| Chennai | 110 | 10.100.2.1 | 10.100.2.2 | APAC |
| London | 200 | 10.100.3.1 | 10.100.3.2 | EMEA |
| Frankfurt | 210 | 10.100.4.1 | 10.100.4.2 | EMEA |
| New Jersey | 300 | 10.100.5.1 | 10.100.5.2 | Americas |
| Dallas | 310 | 10.100.6.1 | 10.100.6.2 | Americas |
| Bangalore | 120 | 10.100.7.1 | - | APAC |
| Delhi | 130 | 10.100.8.1 | - | APAC |
| Noida | 140 | 10.100.9.1 | - | APAC |
Document Revision History¶
| Version | Date | Author | Changes |
|---|---|---|---|
| 1.0 | January 2026 | Network Architecture Team | Initial detailed design |
End of Abhavtech SD-WAN Detailed Capacity Planning & Traffic Flow Design
This document contains actual deployment data for Abhavtech.com's 9-site SD-WAN implementation and should be treated as confidential.