Home » Network Capacity Planning: Bandwidth Estimation, Growth Forecasting, Bottleneck Analysis และ Upgrade Strategy
Network Capacity Planning: Bandwidth Estimation, Growth Forecasting, Bottleneck Analysis และ Upgrade Strategy
Network Capacity Planning: Bandwidth Estimation, Growth Forecasting, Bottleneck Analysis และ Upgrade Strategy
Network Capacity Planning วางแผนให้ network มี capacity เพียงพอสำหรับปัจจุบันและอนาคต Bandwidth Estimation ประเมินความต้องการ bandwidth ปัจจุบัน, Growth Forecasting คาดการณ์การเติบโต, Bottleneck Analysis หาจุดคอขวด, Upgrade Strategy วางแผนการอัปเกรดอย่างคุ้มค่า
Network capacity planning เป็น ศิลปะของการสมดุลระหว่าง “พอ” กับ “ไม่เกิน”: under-provisioning → congestion, poor user experience, business impact | over-provisioning → wasted budget (bandwidth ที่ไม่ใช้ก็ต้องจ่ายเงิน) Rule of thumb: plan for 2-3x current peak usage → upgrade เมื่อ sustained utilization > 70% ของ link capacity องค์กรที่ไม่ทำ capacity planning มักประสบปัญหา “network ช้า” ซ้ำๆ
Bandwidth Estimation
| Application |
Bandwidth Per User/Session |
Concurrent Factor |
| VoIP (G.711) |
100 kbps per call (bidirectional) |
10-20% of users at peak |
| Video Conference (HD) |
2-4 Mbps per session |
5-15% of users at peak |
| Video Conference (4K) |
8-15 Mbps per session |
1-5% of users |
| Web Browsing |
1-5 Mbps (bursty) |
50-70% of users active |
| Email |
50-500 kbps (bursty) |
30-50% of users |
| File Transfer/Backup |
10-100 Mbps (sustained) |
Background, schedule off-peak |
| Cloud/SaaS (O365, etc.) |
2-5 Mbps per user |
60-80% of users |
| Streaming Video (Training) |
5-25 Mbps per stream |
5-10% of users |
Capacity Calculation Formula
| Step |
Formula |
Example (500 users) |
| 1. Per-App Bandwidth |
Users × Concurrent% × BW per session |
VoIP: 500 × 15% × 100kbps = 7.5 Mbps |
| 2. Sum All Apps |
ΣApp₁ + App₂ + … + Appₙ |
VoIP 7.5 + Video 60 + Web 175 + Cloud 200 = ~442 Mbps |
| 3. Add Overhead |
Total × 1.2 (20% protocol overhead) |
442 × 1.2 = ~530 Mbps |
| 4. Growth Factor |
Total × (1 + annual growth%)^years |
530 × (1.3)² = ~896 Mbps (30% growth, 2 years) |
| 5. Headroom |
Design for 70% utilization max |
896 / 0.7 = ~1.28 Gbps → provision 1 Gbps or 2 Gbps link |
Monitoring & Baseline
| Metric |
Tool |
Threshold |
| Link Utilization |
SNMP (interface counters), PRTG, LibreNMS, Zabbix |
Sustained > 70% = plan upgrade |
| Peak vs Average |
95th percentile measurement (industry standard billing) |
95th percentile > 80% = upgrade soon |
| Application Mix |
NetFlow/IPFIX analysis, ntopng, Kentik |
Identify top talkers, unexpected applications |
| Latency/Jitter |
IP SLA probes, ThousandEyes, ping monitoring |
Latency increase during peak = congestion indicator |
| Error Rates |
Interface error counters (CRC, drops, discards) |
Any drops = potential capacity or hardware issue |
| QoS Queue Drops |
QoS statistics on routers/switches |
Drops in priority queues = serious capacity issue |
Bottleneck Analysis
| Common Bottleneck |
Symptoms |
Solution |
| WAN Link |
High utilization, slow cloud apps, poor video quality |
Upgrade bandwidth, SD-WAN (use multiple links), QoS prioritization |
| Internet Breakout |
Slow SaaS, high latency to cloud, proxy overloaded |
Local internet breakout at branch (SD-WAN), upgrade proxy capacity |
| Firewall Throughput |
High CPU on firewall, connection table full, packet drops |
Upgrade firewall (higher throughput), review rules, offload inspection |
| Switch Uplinks |
Access-to-distribution oversubscribed, intermittent slowness |
Add uplink ports (LAG), upgrade to 10G/25G uplinks |
| Wireless Capacity |
Many users per AP, high channel utilization, slow WiFi |
Add APs (smaller cells), use 5/6 GHz, WiFi 6/6E upgrade |
| Server NIC |
Server network interface saturated |
NIC teaming/bonding, upgrade to 10G/25G NIC |
Growth Forecasting
| Factor |
Impact |
Typical Growth |
| User Growth |
More users = more bandwidth |
5-15% per year (depends on organization) |
| Application Adoption |
New apps (video, cloud migration) = step change in bandwidth |
20-50% increase when adopting video/cloud |
| IoT Devices |
Each device adds small bandwidth but thousands of devices add up |
30-50% device growth per year |
| Remote Work |
VPN/ZTNA traffic, video conferencing surge |
Post-COVID: 30-50% remote work = 2x VPN capacity needed |
| Resolution/Quality |
HD → 4K video, higher quality media |
2-4x bandwidth per stream every 3-5 years |
| Overall Rule |
Network traffic doubles every 2-3 years (historical average) |
Plan for 30-50% annual growth in total bandwidth |
Upgrade Strategy
| Strategy |
When |
How |
| Quick Win: QoS |
Congestion but budget not ready |
Implement QoS to prioritize critical apps → buy time before upgrade |
| WAN Optimization |
WAN expensive, can’t easily upgrade |
Compression, deduplication, caching (Riverbed, Silver Peak) → 2-5x effective capacity |
| Link Aggregation |
Need more bandwidth, existing links available |
Bond multiple links (LAG/LACP) → incremental capacity increase |
| Speed Upgrade |
Link consistently > 70% utilization |
1G → 10G, 10G → 25G/40G/100G — check optics, cabling compatibility |
| Architecture Change |
Fundamental bottleneck (three-tier → spine-leaf) |
Redesign network topology — major project, biggest impact |
| Cloud/SASE |
Backhauling to DC is the bottleneck |
Direct internet breakout, SASE, cloud-native architecture |
ทิ้งท้าย: Capacity Planning = Proactive, Not Reactive
Network Capacity Planning Estimation: calculate per-app bandwidth × concurrent users × overhead × growth factor → design at 70% max Monitoring: SNMP utilization (sustained > 70% = upgrade), 95th percentile, NetFlow app mix, latency during peak Bottlenecks: WAN link, firewall throughput, switch uplinks, wireless capacity, internet breakout — identify and fix Growth: traffic doubles every 2-3 years, plan 30-50% annual growth, watch step changes (cloud/video adoption) Upgrade: QoS (quick win), WAN optimization (2-5x), link aggregation (incremental), speed upgrade (1G→10G), architecture change Key: proactive capacity planning prevents “network slow” complaints — monitor, forecast, upgrade BEFORE users notice
อ่านเพิ่มเติมเกี่ยวกับ Network Performance Latency Jitter Throughput QoE และ Network Monitoring SNMP NetFlow Prometheus Grafana ที่ siamlancard.com หรือจาก icafeforex.com และ siam2r.com
