Page
VoIP capacity planning: size voice for busy periods and failure
VoIP capacity is about concurrent calls on the constrained part of the path, not the total number of extensions. A site with 200 users may need little voice bandwidth most of the day and still fail during a busy hour, WAN outage or large meeting when calls concentrate on one link or Wi-Fi cell.
Start with a concurrency estimate
Use call records, contact-centre reporting, reception workload and business events to estimate normal and peak simultaneous calls. Then ask what changes during a degraded state: does another site divert calls here, do users fall back to softphones, or does a carrier route failover traffic through the same Internet connection?
| Input | Why it matters |
|---|---|
| Peak simultaneous calls | Determines the media streams that must fit at once |
| Codec and packet interval | Changes payload size, overhead and quality trade-offs |
| Call direction | One call can create media in both directions across different links |
| Other traffic | Backups, video and cloud sync compete for queues and airtime |
| Failure mode | Secondary paths may be slower or carry more calls than the primary path |
Include real packet overhead
Codec bit rate alone is not a bandwidth plan. RTP is carried inside UDP, IP and a link-layer frame, and encryption or tunnelling can add further overhead. Size both directions, allow headroom for bursts and signalling, and use the actual codec/packetisation settings expected in the service. Where values affect a procurement or circuit decision, measure a representative call rather than relying on a generic rule of thumb.
Treat Wi-Fi as shared airtime
Wi-Fi capacity depends on signal quality, contention, retries, roaming and the mix of client devices. A fast Internet circuit cannot compensate for a crowded or weak wireless segment. Test voice where users stand or move, during real activity, and watch for retry rates and roaming behaviour as well as throughput.
Protect the path with QoS
QoS can prioritise voice where the organisation controls the queue; it cannot create capacity on an already saturated external path. Mark and trust traffic only within a deliberate policy, provide a low-latency queue with sensible limits, and make sure large transfers are managed. Confirm the policy survives switches, wireless infrastructure, VPNs and WAN edges.
Validate the failure state
Test the secondary circuit, backup tunnel or mobile fallback under a realistic call load. Verify not only that calls connect, but that audio remains usable and the monitoring system identifies the degraded condition. Document the maximum supported call load in the fallback state so staff know when to divert or limit activity.
Voice codecs
€” Understand codec choice and trade-offs.
RTP port ranges
€” Plan media paths through firewalls.
QoS and call quality
€” Measure loss, jitter, latency and congestion.
Wi-Fi voice troubleshooting
€” Test airtime, coverage and roaming.
Voice quality testing
€” Turn assumptions into repeatable measurements.
Capacity figures are design inputs, not guarantees. Confirm actual codec behaviour, network overhead, carrier limits and traffic patterns in the intended environment.