When to move recording onto a dedicated gateway server

Most VICIdial deployments write recordings directly on the dialer server. That works well at modest volumes. But at higher call rates, or when compliance demands true Stereo recording with clean channel separation, a dedicated gateway server starts paying for itself quickly.

The three reasons teams move to a gateway

1. Disk I/O competition

Call recording is write-heavy. When your dialer server is simultaneously running Asterisk, the VICIdial daemon, MySQL, and MixMonitor for every active call, the disk becomes the bottleneck first. Moving recording I/O to a separate box lets each server do one job well.

2. True stereo channel separation

The VICIrecGateway AGI uses MixMonitor's r() and t() flags to write the carrier audio and the dialer audio into separate file paths before they are mixed. On the main dialer, by the time you hear both sides in a single mono file, the signals are already blended. The gateway captures them before the mix happens.

3. Compliance isolation

Some Recording retention and PCI-related frameworks require that recording infrastructure be separate from the system that processes transactions. A gateway server satisfies that isolation requirement without changing how agents work.

When the gateway is probably not worth it

• Fewer than 30 concurrent calls: the dialer server handles recording I/O without strain at this volume.
• Mono recordings are fine for your use case: if you only need playback or basic QA, the extra infrastructure cost is not justified.
• No transcription or voice-analytics pipeline: stereo pays off most when a downstream tool can exploit the channel separation.

Decision flow

flowchart TD
  A[Recording needed] --> B{More than 30 concurrent calls?}
  B -->|N| C[Record on main dialer]
  B -->|Y| D{Need stereo or compliance isolation?}
  D -->|N| E[Add dedicated recording disk on dialer]
  D -->|Y| F[Deploy VICIrecGateway server]
  F --> G[stereo-rec-inbound context]
  F --> H[stereo-rec-outbound context]
  G --> I[CARRIER_FILENAME and DIALER_FILENAME written separately]

What the gateway server needs

The VICIrecGateway runs on a standard Asterisk install with PJSIP configured. Your trunk provider sends calls to the gateway's inbound DID, the gateway runs the AGI, fires MixMonitor in the stereo-rec-inbound context, then dials forward to your main VICIdial server via a PJSIP peer. It injects the X-VICIrecGatewayID header so the dialer can match the recording files to the call record.

For outbound calls, the dialer routes through stereo-rec-outbound, which swaps the r() and t() channel assignments so that CARRIER_FILENAME always captures the far-end audio regardless of call direction.

**Heads up:** Every active call passes through the gateway. Network latency between the gateway and your main dialer adds to call setup time and can affect audio quality if the link is slow or lossy. Run the gateway on a server in the same data center as your dialer, not across a WAN.

Storage planning on the gateway

Because the gateway writes two files per call instead of one, your storage budget roughly doubles compared to mono recording at the same volume. Plan accordingly and apply the same Recording retention lifecycle policies you use on the main dialer — archive to cold storage, prune by age, or ship to object storage on a schedule.

For broader recording architecture, see VICIdial call recording explained. For a technical breakdown of what VICIrecGateway does inside the dialplan, see the VICIrecGateway AGI explained.

VICIfast can include a gateway recording server in your managed deployment. See VICIfast pricing to discuss your recording architecture.