How Many SIP Trunks Do I Need? Sizing Guide

This is a common question from operators migrating from PRI to SIP, and the answer turns on a difference in how the two technologies define capacity. On PRI, the question was essential — each T1 gave you exactly 23 channels, so you needed to calculate your peak concurrent calls and buy enough T1s to cover it. On SIP, the question is usually irrelevant, because the concept of “how many trunks” does not map to SIP the way it mapped to PRI.

This guide explains the actual capacity planning question for SIP trunking and gives you the math to size your operation correctly. SIPNEX is an FCC-licensed carrier with unlimited concurrent channels — the concept of buying “enough trunks” does not apply on our platform. But bandwidth planning, codec selection, server capacity, and calls-per-second (CPS) limits still matter, and that is what this guide covers.

Why “how many trunks” is the wrong question

On PRI, a trunk was a physical circuit with a fixed number of channels. One T1 = one PRI = 23 channels. The trunk was the unit of capacity. Buying more trunks meant buying more capacity.

On SIP, a trunk is a logical connection between your PBX and the carrier — defined by a SIP registration (or IP-based authentication) to the carrier’s proxy. On carriers that don’t cap channels, like SIPNEX, that single logical connection carries as many simultaneous calls as your bandwidth and servers allow. You do not need multiple trunks for more capacity. You need one trunk with enough bandwidth to carry your peak concurrent call volume.

The correct question is not “how many SIP trunks do I need?” It is “how many concurrent calls will my operation generate at peak, and do I have the bandwidth and server resources to handle them?”

Some carriers — particularly resellers — do impose per-trunk or per-account concurrent channel limits (10, 25, 50, or 100 channels). On those carriers, the PRI-era question is relevant: you need to buy enough channel capacity to cover your peak concurrency. That cap is a commercial term, not a technical limit of SIP. SIPNEX doesn’t cap concurrent channels, so on our trunks capacity comes down to your bandwidth, your servers, and the CPS limit covered below.

Calculate your channels and bandwidth

The calculator below runs the same math this guide walks through by hand — peak concurrent calls for your operation type, a 25% burst headroom, and the bandwidth those channels consume at your codec’s rule-of-thumb rate.

The concurrent call calculation

The fundamental calculation is simple: how many phone calls will be active simultaneously at your busiest moment?

For inbound operations (customer service, support lines): count the maximum number of calls in queue plus calls being handled by agents at the busiest hour. If you have 30 agents and your queue holds up to 20 callers during peak hours, your peak concurrency is approximately 50 simultaneous calls.

For outbound manual/preview dialing: peak concurrency roughly equals the number of agents actively dialing. 50 agents on preview mode = approximately 50 concurrent calls.

For outbound predictive dialing: peak concurrency equals agents multiplied by the dial ratio. The dial ratio is the number of simultaneous outbound call attempts per available agent — it depends on your answer rate and your abandon rate target. Common ratios:

• Conservative (1-2% abandon target): 2:1 to 3:1 ratio
• Standard (2-3% abandon target): 3:1 to 4:1 ratio
• Aggressive (approaching 3% abandon limit): 4:1 to 5:1 ratio

50 agents × 4:1 ratio = 200 concurrent channels at peak. 100 agents × 3:1 = 300 concurrent channels. Add 20-30% headroom for burst periods (when answer rates suddenly drop and the dialer compensates with more aggressive pacing).

For VICIdial specifically: check your real-time campaign dashboard during peak hours. VICIdial shows active channels in the system status display. Your peak concurrent channel count from historical data is the number to plan around, plus 20-30% buffer.

Calls per second: the other carrier-side limit

Uncapped channels don’t mean unlimited call attempts. Most carriers also set a calls-per-second (CPS) limit per account: how many new calls you can launch each second. Inbound and manual dialing rarely come near it. A predictive dialer launches attempts in bursts, though, and a CPS ceiling below that burst rate rejects or delays attempts even while plenty of channels sit idle.

Your dialer has a ceiling of its own. The ViciBox dimensioning guide (docs.vicibox.com) recommends about 5 CPS per VICIdial telephony server. Ask any carrier, SIPNEX included, to confirm in writing the CPS your account can send before a large campaign launches.

The bandwidth calculation

Once you know your peak concurrent call count, calculate the bandwidth requirement based on your codec selection.

G.711u (ulaw): The highest quality codec and the default for most VoIP deployments. At the standard 20 ms packet size, each call uses about 87 kbps in each direction on Ethernet: 64 kbps of audio, 16 kbps of IP/UDP/RTP headers, and about 7 kbps of Ethernet framing (Cisco’s per-call bandwidth tables list 87.2 kbps). At 200 concurrent calls: 200 × 87 kbps ≈ 17.4 Mbps upstream and 17.4 Mbps downstream, or about 35 Mbps of voice traffic in total.

G.729: A compressed codec that reduces bandwidth consumption at the cost of slight audio quality reduction. Each call uses about 31 kbps in each direction on Ethernet at 20 ms packets (Cisco lists 31.2 kbps), so plan on 32 kbps. At 200 concurrent calls: 200 × 32 kbps = 6.4 Mbps each direction. Total: 12.8 Mbps. G.729 is appropriate when bandwidth is limited (satellite links, rural connections), but G.711 is preferred for call center operations where audio quality directly impacts agent effectiveness and customer experience.

Opus: A modern, adaptive codec that adjusts quality based on available bandwidth. Its bitrate can range from 6 kbps to 510 kbps (RFC 6716), and voice calls use the low end of that range. Opus is increasingly supported but not yet universal in carrier environments.

Rule of thumb: plan for 100 kbps per concurrent call when using G.711. This rounds up from about 87 kbps to account for signaling overhead (SIP messages), RTCP (control protocol), and network inefficiencies. 200 concurrent calls = 20 Mbps dedicated to voice. 500 concurrent calls = 50 Mbps.

Quality matters more than raw speed. A 100 Mbps connection with 5% packet loss and 200ms jitter will produce worse call quality than a 20 Mbps connection with 0.1% packet loss and 20ms jitter. For voice traffic, prioritize connection quality: low jitter (under 30ms), low packet loss (under 1%), and low latency (under 150ms one-way). If you share your internet connection with data traffic, implement QoS (Quality of Service) to prioritize voice packets. If your operation is large enough, dedicate a separate internet connection for voice.

Server capacity considerations

Your SIP trunk’s capacity is also limited by your PBX or dialer server’s processing power.

Asterisk/VICIdial capacity depends less on core count than on how much channel activity Asterisk can handle. The same ViciBox dimensioning guide (ViciBox is the official VICIdial installation image) plans a telephony server at about 25 agents and 125 carrier channels for predictive dialing, or about 50 agents and 250 channels for inbound-only or light predictive work. Extra cores add little: the guide says a quad-core server may run into trouble near 25 agents while an eight-core one might reach only about 30.

Features and codecs move those numbers. With call recording, ViciBox calls for its recommended hardware spec rather than its minimum. G.711 needs less CPU than G.729, because Asterisk doesn’t have to compress and decompress it. Answering Machine Detection (AMD), which most predictive campaigns run on every call, adds load of its own, so measure it on your own campaigns rather than planning around a generic percentage.

Call recording adds disk I/O requirements. A call recorded as G.711 audio (64 kbps) writes about 480 KB per minute, so 100 concurrent recorded calls add roughly 6.4 Mbps, about 800 KB per second, of sustained writes. Ensure your storage can handle that throughput during peak hours and that you have adequate disk space for retention (a 100-agent operation recording 8 hours of calls per agent per day generates roughly 23 GB of audio daily before any compression).

Plan servers alongside bandwidth. By the ViciBox figures, one predictive telephony server’s 125 channels need only about 12.5 Mbps of G.711 bandwidth each way at the 100 kbps rule of thumb, so on VICIdial the telephony server count is often the first limit you plan around. Past one server, VICIdial scales as a cluster: a database server, web servers, and as many telephony servers as your peak channel count requires.

Capacity planning checklist

Use this to size your SIP trunking deployment:

Calculate peak concurrent channels: agents × dial ratio × 1.25 (25% burst buffer). Example: 75 agents × 3.5 ratio × 1.25 = 328 concurrent channels.

Calculate bandwidth requirement: concurrent channels × 100 kbps. Example: 328 × 100 kbps = 32.8 Mbps dedicated to voice.

Verify internet connection quality: run a VoIP-specific speed test that measures jitter, packet loss, and latency, not just download/upload speed. Target: jitter under 30ms, packet loss under 1%, latency under 150ms.

Verify server capacity: divide peak channels by what one telephony server carries. ViciBox plans about 125 channels per server for predictive dialing, so the 328-channel example needs about three telephony servers. Confirm with load tests on your own hardware.

Verify carrier limits: confirm your carrier’s concurrent channel cap (SIPNEX has none) and the CPS your account can send. If either is below your peak requirement, raise it or switch carriers.

Plan for growth: if your operation is growing, size for projected peak, not current peak. Adding bandwidth and server capacity is straightforward. Adding carrier capacity is only an issue if your carrier caps channels or your account’s CPS is too low for the new volume.

Frequently asked questions

How many SIP trunks do I need for my business?

On a carrier like SIPNEX with unlimited concurrent channels, you need one SIP trunk (one logical connection). That trunk handles as many simultaneous calls as your bandwidth and servers can support, subject to any calls-per-second (CPS) limit on your account. The real question is how many concurrent calls your operation generates at peak — agents times dial ratio for predictive dialing, or agent count for manual dialing — and whether your internet bandwidth can carry that volume at about 100 kbps per concurrent G.711 call. On carriers that cap concurrent channels, buy enough channels to cover peak concurrency plus a 25% burst buffer.

Can I start small and scale up?

Yes, and this is one of the primary advantages of SIP over PRI. On PRI, scaling up meant ordering and installing a new physical T1 circuit — a process that took weeks and committed you to the cost of 23 additional channels whether you used them or not. On SIP, scaling up means making more calls. If you start with 10 agents and grow to 100, you do not need to upgrade your trunk — you need to upgrade your bandwidth and possibly your server. On SIPNEX there are no channel limits to adjust as you grow. The carrier-side item to revisit is calls per second: confirm the CPS your account can send before a much larger predictive operation goes live.

What is the difference between SIP channels and SIP trunks?

A SIP trunk is the logical connection between your phone system and the carrier — defined by your SIP registration or IP authentication to the carrier’s proxy. A SIP channel is a single simultaneous call on that trunk. One trunk can carry many channels. On PRI, these concepts were equivalent: one trunk (T1) = 23 channels. On SIP, they are decoupled: on carriers that don’t cap channels, like SIPNEX, one trunk carries as many channels as your bandwidth and servers support. Carriers that sell SIP channels separately are pricing capacity the way PRI did; that limit is commercial, not technical.

---

SIPNEX trunks don’t cap concurrent channels. To check your sizing, including the CPS your campaigns need, contact SIPNEX with your channel and bandwidth estimates.