Network Capacity Planning: Methodology and Tools

Proactive capacity planning prevents outages before they happen. This is the methodology for enterprise networks.

Step 1: Establish Baselines

Collect at least 90 days of data before planning. Key metrics:

TEXT

Per Interface:
- Inbound/Outbound utilization (avg, 95th percentile, peak)
- Error rate (input/output errors per hour)
- Drops (queue drops)

Per Device:
- CPU utilization (avg, peak)
- Memory utilization
- BGP prefix counts (if applicable)

Per Site:
- WAN utilization (daily pattern)
- User count trend

Traffic Analysis with NTOPNG

BASH

# Install ntopng
apt install ntopng

# Configure to monitor interface
# /etc/ntopng/ntopng.conf
--interface=ether1
--http-port=3000
--community

systemctl enable --now ntopng

Utilization Thresholds

Threshold	Action
50%	Plan capacity upgrade
70%	Begin procurement process
80%	Implement QoS/traffic shaping
90%	Emergency upgrade (if no plan)

Note: Use 95th percentile, not average. Average hides bursty traffic.

Growth Modeling

PYTHON

import numpy as np

# Monthly bandwidth usage (GB)
history = [120, 135, 142, 158, 170, 185]

# Linear regression forecast
months = np.arange(len(history))
z = np.polyfit(months, history, 1)
p = np.poly1d(z)

for m in range(6, 18):  # Next 12 months
    forecast = p(m)
    print(f"Month {m}: {forecast:.0f} GB")

WAN Capacity Planning Worksheet

TEXT

Current WAN speed: 100 Mbps
Current 95th percentile utilization: 72 Mbps (72%)
Monthly growth rate: 8%

Month 1: 77.8 Mbps (78%)
Month 2: 84.0 Mbps (84%) ← Needs upgrade
Month 3: 90.7 Mbps (91%) ← Emergency territory

Procurement Timeline

Work backwards from the capacity threshold:

Month 0: Order placed (today)
Month 1-2: Procurement/vendor lead time
Month 2-3: Installation and testing
Month 3: Live at new capacity

Start process when utilization hits 50%, not 90%.

Reporting Template

Produce monthly capacity reports with:

Current utilization by site/interface
Trend chart (6 months)
Forecast to 80% threshold date
Recommended action and timeline