CCDE SD-WAN Design Part 14: Internet Circuit Upgrade Requirements Before MPLS Migration

CCDE Enterprise Case Study Part 14 – SD-WAN Internet Circuit Capacity Planning

As Jacobs continues evaluating SD-WAN migration strategies, one of the most important architectural considerations is whether the existing Internet infrastructure can realistically support migration of WAN traffic from MPLS toward Internet-based SD-WAN overlays.

This question is fundamentally about:

• Bandwidth engineering
• Capacity planning
• Traffic forecasting
• Scalability analysis
• Internet edge architecture

🎯 Correct Answer

The correct answer is:

c. Upgrade Internet access circuits to 10Gbps in the DCs.

The current 1Gbps Internet circuits at Jacobs and Toolmate are already approaching saturation.

Migrating additional SD-WAN traffic from MPLS onto Internet transport would overload the existing circuits.

Table of Contents

• 1. Understanding the Question
• 2. Why Option C Is Correct
• 3. WAN Capacity Planning Fundamentals
• 4. ISP Utilization Graph Analysis
• 5. MPLS to Internet Traffic Migration
• 6. WAN Engineering Mathematics
• 7. SD-WAN Internet Architecture
• 8. Why the Other Answers Are Wrong
• 9. CLI and Monitoring Examples
• 10. Machine Learning and Traffic Forecasting
• 11. Related Articles

1. Understanding the Question

The key architectural problem is:

$$ Can\\ Existing\\ Internet\\ Circuits\\ Support\\ SD\\text{-}WAN\\ Migration? $$

Jacobs currently uses:

• MPLS for primary WAN transport
• Internet mainly for Internet-bound traffic
• Centralized Internet breakout

If SD-WAN is introduced:

• Branch-to-DC traffic may traverse Internet overlays
• Cloud traffic may increase dramatically
• Hybrid WAN transport becomes active
• Encrypted tunnel overhead increases utilization

This fundamentally changes WAN traffic behavior.

2. Why Option C Is Correct

Current Internet Links Are Already Highly Utilized

The supplied utilization graphs show:

• Jacobs ISP CE1 nearing 1Gbps capacity
• Toolmate ISP utilization also nearing saturation

This means:

$$ Existing\\ Capacity \approx Existing\\ Demand $$

Adding SD-WAN transport traffic would cause:

• Congestion
• Packet loss
• High latency
• Jitter
• Application degradation

Bandwidth Utilization Formula

$$ Utilization = \frac{Traffic}{Available\\ Bandwidth} \times 100 $$

If:

$$ Traffic > Available\\ Bandwidth $$

then:

$$ Congestion = True $$

Why 10Gbps Makes Sense

Jacobs currently has:

• Hundreds of branches
• VoIP traffic
• Cloud applications
• Jaystore ERP traffic
• Future SaaS expansion

A move toward SD-WAN dramatically increases Internet dependency.

Therefore:

$$ Internet = Primary\\ Enterprise\\ Transport $$

not simply:

$$ Internet = Guest\\ Access $$

3. WAN Capacity Planning Fundamentals

Capacity planning is one of the most important enterprise design disciplines.

Enterprise WAN Capacity Variables

Variable	Impact
Users	Traffic growth
Applications	Bandwidth demand
Cloud adoption	Internet dependency
Video/VoIP	Real-time sensitivity
Encryption overhead	Tunnel expansion

Forecasting Formula

$$ Future\\ Capacity = Current\\ Demand \times Growth\\ Factor $$

For example:

$$ 1Gbps \times 4 = 4Gbps $$

If SD-WAN quadruples Internet transport usage, the existing links fail immediately.

4. ISP Utilization Graph Analysis

Jacobs ISP CE1 Analysis

• Peak utilization approaches 0.9Gbps
• Average utilization remains high
• Traffic volatility exists

This indicates:

$$ Limited\\ Remaining\\ Headroom $$

Toolmate ISP Analysis

• Inbound traffic increasing steadily
• Peak inbound nearing line capacity
• Outbound traffic highly variable

This strongly suggests:

$$ Cloud\\ and\\ SaaS\\ Growth $$

is already occurring.

💡 Key Design Insight

Enterprise architects must always design for:

$$ Peak\\ Demand $$

not:

$$ Average\\ Demand $$

5. MPLS to Internet Traffic Migration

Traditional MPLS WANs typically isolate:

• Private WAN traffic
• Internet traffic

SD-WAN merges these into:

$$ Hybrid\\ WAN\\ Transport $$

Before SD-WAN

Traffic Type	Transport
Internal WAN traffic	MPLS
Internet traffic	Centralized breakout

After SD-WAN

Traffic Type	Transport
Internal WAN traffic	Internet + MPLS
SaaS traffic	Direct Internet
Cloud applications	Hybrid transport

This dramatically changes:

$$ Internet\\ Bandwidth\\ Consumption $$

6. WAN Engineering Mathematics

Oversubscription Ratio

$$ Oversubscription = \frac{Potential\\ Demand}{Available\\ Bandwidth} $$

Tunnel Overhead Formula

$$ Effective\\ Payload = Bandwidth - Encryption\\ Overhead $$

Latency Growth Under Congestion

$$ Latency \propto Queue\\ Depth $$

Packet Loss Probability

$$ Loss \propto Buffer\\ Exhaustion $$

VoIP MOS Relationship

$$ MOS \downarrow \quad when \quad Latency + Jitter + Loss \uparrow $$

7. SD-WAN Internet Architecture

Modern SD-WAN solutions commonly use:

• MPLS
• Broadband Internet
• 5G/LTE

simultaneously.

Traffic engineering becomes:

$$ Application\\ Aware $$

Typical SD-WAN Overlay

Branch Edge
   |
Internet/MPLS
   |
Encrypted Tunnel
   |
DC SD-WAN Edge

Critical Enterprise Reality

When Internet becomes:

$$ Mission\\ Critical $$

Internet circuits must be engineered like:

$$ Carrier\\ Grade\\ Infrastructure $$

8. Why the Other Answers Are Wrong

Option A – Firewall Upgrade

This is suboptimal because:

• No evidence firewall throughput is exhausted
• SD-WAN edges usually terminate tunnels
• Firewalls may simply pass encrypted traffic

Option B – Allow SD-WAN Flows

This is merely:

$$ Implementation $$

not:

$$ Architecture $$

Firewall rule creation does not solve:

$$ Bandwidth\\ Saturation $$

Option D – Routing Protocol Selection

OMP selection is implementation detail.

The major issue is:

$$ Physical\\ Capacity $$

not:

$$ Routing\\ Mechanics $$

9. CLI and Monitoring Examples

Example – Interface Utilization Monitoring

show interface GigabitEthernet0/0

  5 minute input rate 850000000 bits/sec
  5 minute output rate 790000000 bits/sec

Example – SD-WAN Tunnel Statistics

show sdwan tunnel statistics

Example – QoS Policy

policy-map WAN-QOS

 class VOICE

  priority percent 30

 class BUSINESS-CRITICAL

  bandwidth percent 40

Why QoS Still Matters in SD-WAN

SD-WAN improves path selection but does not eliminate congestion.

QoS still remains critical during:

• Circuit oversubscription
• WAN brownouts
• Application bursts
• Cloud synchronization spikes

10. Machine Learning and Traffic Forecasting

Modern enterprises increasingly use:

• AI-driven telemetry
• Traffic prediction
• Anomaly detection
• Forecast modeling

to proactively scale WAN infrastructure.

Useful Machine Learning Concepts

• Time Series Forecasting Beginners Guide
• Stationary vs Nonstationary Data
• How to Evaluate and Ensure Your Data

These concepts directly apply to:

$$ Network\\ Traffic\\ Forecasting $$

11. Related Articles

CCDE Enterprise Case Study Series

• Part 1 – Enterprise Architecture
• Part 2 – Business Challenges
• Part 3 – Scalability Analysis
• Part 4 – MPLS Design
• Part 5 – MPLS Traffic Engineering
• Part 6 – DMVPN Architecture
• Part 7 – Enterprise WAN Evolution
• Part 8 – SD-WAN Design
• Part 9 – SD-WAN Security
• Part 10 – WAN Optimization
• Part 11 – Self-Managed SD-WAN
• Part 12 – SD-WAN Planes Explained
• Part 13 – WAN Path Engineering
• CCDE SD-WAN Design Part 15: 5G, DIA, Firewall Placement, and Branch Security Architecture

CCIE and Enterprise WAN Articles

• Understanding BGP Path Selection
• Enterprise QoS Design Best Practices
• Complete MPLS QoS Configuration Lab
• Enterprise Redundancy and High Availability Design

Final Conclusion

The most critical issue Jacobs must address before migrating toward Internet-based SD-WAN transport is:

$$ Internet\\ Circuit\\ Capacity $$

The current ISP utilization graphs clearly demonstrate:

• Minimal remaining bandwidth headroom
• Growing traffic demand
• Potential congestion risk

Without upgrading the Internet circuits:

$$ SD\\text{-}WAN\\ Migration \rightarrow Performance\\ Failure $$

Enterprise architects must always remember:

$$ Architecture\\ Success = Scalable\\ Physical\\ Infrastructure $$