A VoIP gateway handles exactly as many concurrent calls as it has ports: a 16-port gateway carries 16 simultaneous conversations, and each new call waits for a port to free up. The published port count is the ceiling, and the practical concurrency is usually lower, because codecs, signaling, SIM health, and network bandwidth each take their share of the real capacity.
This guide explains the port equation, the factors that reduce real concurrency, and the sizing formula that turns a call pattern into a VoIP gateway specification.
The Port Equation
One port equals one active call channel for the duration of the conversation. When a caller dials out, a port is taken at setup and released at hangup, and the gateway can only start a new call when a port is free; that is the entire capacity model.
The equation is deceptively simple because the call duration matters as much as the call count. A gateway can carry hundreds of calls in a day across a few ports, because each port cycles through many short conversations; concurrency is about how many overlap at one moment, not how many happen in total.
The planning number is the peak of simultaneous calls, measured in the busiest five minutes, not the average hour. A reception desk that sees six calls an hour may never exceed two concurrent, while a call center with the same daily volume may peak at twelve.
The measurement method matters more than the math: pull the call detail records from the PBX, count active calls in five-minute buckets across the busiest month, and note the date and hour of the peak. That record is the specification, and it is the same number the supplier needs to size the unit.
Short calls and long calls change the picture: a call center with two-minute calls cycles ports quickly, while a support line with twenty-minute calls holds ports for long stretches. The average duration belongs in the record alongside the peak, because it explains why the same daily volume needs different port counts.
The distinction between ports and channels matters on multi-SIM gateways: a port is the call channel, and a SIM is the carrier connection behind it, and the two counts are configured independently. Confirm both numbers, because a gateway with more ports than usable SIMs delivers the smaller number.
What Reduces Real Concurrency
The first reducer is the codec and bandwidth pair: each concurrent call consumes uplink and downlink bandwidth according to the codec, and a saturated network link makes calls degrade or fail before the port ceiling is reached. The port count assumes bandwidth exists for every port.
The second reducer is SIM and carrier behavior on GSM gateways: concurrent calls share the SIM pool's carrier limits, and a throttled SIM or weak signal turns a port into a failing channel. Sixteen ports with two healthy SIMs deliver far less than sixteen independent calls.
The third reducer is signaling load and system capacity: registration refreshes, call setup, and media handling all consume the gateway's processor, and a gateway near its port ceiling on a busy signaling day can drop registrations or delay setup even with free ports.
Quality of service settings are the bandwidth mitigation: marking voice traffic with priority on the office router keeps a large download or backup from starving the calls. Without QoS, the gateway can be perfectly configured and still drop audio during peak internet use.
Transcoding is the codec-level cost: when the gateway must convert between codecs for every call, it uses more processor and can add latency and echo. Prefer a common codec across the system to reduce transcoding, and confirm the gateway's codec handling for the ones you cannot avoid.
Network conditions outside the office count too: the path between the gateway and the carrier or provider adds jitter and loss, especially on satellite or congested links. Test the path during busy hours, because the office's local bandwidth is only half of the media path.
Monitoring the reducers is part of capacity management: track bandwidth utilization, SIM health, and call setup times alongside port usage, because each reducer announces itself in its own metric before calls fail. The dashboard that shows all four is the capacity instrument.
The gateway's own limits are worth checking in the datasheet: maximum registrations, sessions, and throughput are separate specifications from the port count. A model that meets the port number but not the session limit is a model that under-delivers in a multi-PBX deployment.
| Factor | Effect | Mitigation |
|---|---|---|
| Codec and bandwidth | Degrades audio, fails calls | QoS, sufficient uplink |
| SIM health | Failing channels | Carrier diversity, monitoring |
| Signaling load | Setup delays, drops | Headroom in port choice |
| Media handling | CPU pressure at peak | Size above the measured peak |
The table is the capacity reality check: every row is a reason the practical concurrency sits below the published number, and every row has a mitigation that belongs in the deployment plan.
The Sizing Formula
The formula has three steps: measure the peak of concurrent external calls from call records or a measurement week, add headroom of 1.5 to 2 times, and select the port tier that covers the result. The headroom absorbs retries, transfers, and unexpected peaks.
Worked example: a business measures a peak of 10 concurrent external calls during the morning rush. With 1.5 times headroom the target is 15, which points to a 16-port gateway; with a growth plan that adds a second team, the 32-port tier is the safer choice.
The formula should be re-run after major changes: a new campaign, a reorganized front desk, or a seasonal peak all move the number. The record of measurements, tier choices, and re-runs is what makes the gateway capacity a managed number rather than a guess.
The formula works for growth as well: add the planned extension count and the expected call behavior to the record, and re-run the formula with the projected peak. A gateway bought with growth headroom avoids the mid-life replacement that costs more than the original purchase.
Load testing is the final proof: run a scripted set of concurrent calls against the configured gateway and watch where quality degrades. The test converts the formula's estimate into a measured ceiling, and it is the evidence a buyer wants before production traffic arrives.
Keep the load-test results with the measurement record, because they are the evidence the next capacity review starts from.
Telarvo Expert Views
The published port count is a ceiling, not a promise: bandwidth, SIM health, and signaling each take a slice of the real capacity. We advise sizing from the measured peak with headroom, and treating the port tier as a managed decision that gets re-measured when the call pattern changes.
— Voice Solutions Engineer, Telarvo Store
Validation note: codec, network, and carrier conditions vary by site; validate capacity with a load test in your environment.
Conclusion
Concurrent call capacity is the port equation with reality attached: one port per active call, reduced by bandwidth, SIM health, and signaling load, and sized from the measured peak with headroom.
Key Takeaways for B2B Buyers
Measure the peak of concurrent calls in five-minute buckets, add 1.5 to 2 times headroom, check bandwidth and SIM health against the port count, and re-measure when the call pattern changes.
Questions to Ask Before Committing
Ask what the published port count assumes about codecs and bandwidth, how SIM health is reported, what the signaling load looks like at the ceiling, and which tier covers your measured peak.
Ask Telarvo Store to size a VoIP gateway from your measured call peak before you commit.
FAQs
Is a port the same as a call?
Yes; each active conversation occupies one port for its duration, and a new call waits for a free port.
How many calls can a gateway handle per day?
Far more than its port count, because ports cycle through many short calls; concurrency is about simultaneous calls, not daily volume.
Why do calls fail below the port ceiling?
Usually bandwidth, SIM health, or signaling load; each consumes real capacity even when ports are free.
Do I need headroom in the port count?
Yes; 1.5 to 2 times the measured peak absorbs retries, transfers, and unexpected traffic.
Does the gateway capacity include conference calls?
Yes; a conference with three participants uses three ports, and transfers can briefly hold two, so count them in the peak measurement.
What is a safe bandwidth figure per call?
It depends on the codec; a common planning figure is 80 to 100 kbps per call in each direction, but confirm with the codec's specification and your network.