Single-server vs clustered VICIdial: when one box is enough

One of the first decisions you make when building a VICIdial system is whether to run everything on a single box or spread the work across several. Most new operators do not need a cluster, and a cluster brings real complexity. This post walks through what each setup gives you so you can pick honestly.

What a single server actually does

On a single-server install, one machine runs everything: the web and database layer, the Asterisk telephony engine that places and bridges calls, and the dialing scripts that load leads and pace your agents. A modest server can comfortably handle dozens of agents and the Concurrent calls needed to keep them busy. For a team under roughly 30 to 40 agents, this is usually the right answer. Fewer moving parts means fewer things to break and one place to look when something goes wrong.

The limit on a single box is rarely agent count alone. It is the combination of dialing rate, recording load, and database writes all landing on the same CPU and disk. When any one of those starts to saturate, you feel it as slow agent screens, delayed lead delivery, or dropped audio.

What a cluster adds

A clustered VICIdial setup splits the workload across multiple machines. The most common move is to add one or more extra dialer boxes, each running its own copy of Asterisk, all pointed at a shared database. This lets you place far more Concurrent calls than one server could and spreads recording disk usage. A second common split is moving the database and web server onto their own dedicated machine so the dialers do nothing but make calls.

In a multi-dialer cluster, VICIdial can use balance dialing to keep a campaign's dial level met even when no agents are logged into a given box, shifting fill calls between dialers by rank. Each dialer connects to your carrier through its own Server trunk, so capacity adds up across boxes. A shared database also opens the door to rough Failover planning, since a second dialer can carry the load if one goes down.

flowchart TD
  DB[Shared database] --> D1[Dialer one Asterisk]
  DB --> D2[Dialer two Asterisk]
  DB --> WEB[Web and admin server]
  D1 -- carrier trunk --> NET[Phone network]
  D2 -- carrier trunk --> NET
  WEB --> AG[Agents log in]
  AG --> D1
  AG --> D2

The tradeoffs

A cluster is more powerful but also more fragile in operational terms. Every box runs its own Keepalive scripts and crontab, every box needs the same software version, and conf files must stay in sync across the fleet. A clock that drifts on one dialer or a carrier credential that is wrong on a single box can cause failures that are hard to trace because the symptom shows up unevenly. You also pay for more servers and more time spent maintaining them.

• Stay single-server while one box keeps your agents busy and CPU and disk have headroom.
• Add a dialer when you run out of call capacity, not when you simply want redundancy.
• Split the database onto its own box when web screens lag while calls are flowing.

This piece is part of our complete guide to installing VICIdial. If you are weighing the work of running your own boxes, the hidden cost of self-hosted VICIdial covers what a cluster really demands of your time.

If you would rather not size, sync, and babysit a cluster yourself, VICIfast provisions a managed VICIdial server in under 40 seconds so you can grow capacity without owning the cluster headache.