Web hosting
Bandwidth capacity planning: size links from real traffic peaks and transfer windows
Bandwidth capacity planning estimates sustained and peak transfer, but the useful metric depends on the service: web response bytes, backup windows, API traffic, software downloads, media, or replication. “Unlimited bandwidth” plan wording does not eliminate network interface, fair-use, CPU, origin, or user-latency constraints.
Measure bytes and peak rate separately
Monthly transfer answers billing/aggregate volume questions; Mbps/Gbps over short intervals answers whether links saturate during peaks. Record both inbound and outbound because backups/imports can differ from customer traffic.
Use percentiles and time-series measurements across representative periods.
Map traffic classes
Separate public web/CDN traffic, API responses, database replication, backups, package updates, uploads, mail, and administrative transfers. Large background transfers can be scheduled or rate-limited independently from latency-sensitive traffic.
Do not serve large media downloads from a web-hosting path without checking storage/CPU/cache architecture.
Estimate growth from actual payloads
Multiply page/API/download response sizes by observed request counts and account for cache/CDN hit rate when relevant. Compression, image optimization and caching can reduce origin bytes, but measure their actual effect on your workload.
For backups, use daily changed bytes and allowed backup window.
Plan headroom and monitoring
Alert on sustained interface utilization, packet drops/retransmissions, transfer-duration overruns, and abrupt usage changes. Upgrade network/service architecture when measured peaks approach the safe operating envelope.
- Track transfer and peak rate.
- Classify traffic types.
- Model measured payloads/growth.
- Reserve headroom for bursts.