Reading VICIdial server graphs for beginners

The first time you open the VICIdial Server Performance Report, you get four wavy graphs and no idea which line means trouble. Good news: there are only four of them, and once you know what each one tracks, the report reads like a story about your shift. Let us walk through them.

First, the setup. This report samples your server every 5 seconds and graphs the result over a time window you choose. The graph section plots four things: the number of processes, the system load, the number of Asterisk channels, and the USER versus SYSTEM processor time. Each one answers a different question.

Graph 1: number of processes

A process is just a running program on the box. VICIdial keeps a set of background programs alive at all times, kept running by a watchdog called the keepalive (see Keepalive). The line should stay roughly flat and rise gently when more agents dial. A sudden cliff or spike is the interesting part, because it usually means a process died or a flood of new work arrived.

Graph 2: system load

System load is a rough gauge of how much work is queued up waiting for the processor. Low and steady is healthy. The shape matters more than any single value: a smooth rise into your busy hour and a fall after is exactly what you want. A line that climbs and stays pinned at the top, or spikes far above its neighbors, is where you look for a problem.

Graph 3: number of Asterisk channels

A channel is one active audio leg through Asterisk, the telephony engine under VICIdial (see Asterisk). Every ringing or connected call uses channels (see Channel). This graph tracks how many concurrent calls were live across the shift (see Concurrent calls). When you dial ahead of your agents to keep them busy, this line climbs higher than the agent count, which is normal for predictive dialing (the mode that places several calls per free agent, see Predictive dialing).

Graph 4: USER vs SYSTEM processor time

Processor time splits into USER (time spent on your applications, like Asterisk and the dialer) and SYSTEM (time the operating system spends on its own plumbing). Whatever is left is IDLE, the breathing room. On a healthy dialer USER dominates. If SYSTEM climbs unusually high, the box is spending its energy on overhead rather than calls, which is worth a closer look.

Here is how the four graphs relate when you read them together:

flowchart LR
  A[More agents and calls] --> B[Channels graph rises]
  B --> C[Processes graph rises]
  C --> D[USER processor time rises]
  D --> E[System load rises]
  E --> F{IDLE still left?}
  F -->|Yes| G[Box is keeping up]
  F -->|No| H[Box is at its limit]

Read them as a chain. Calls drive channels, channels drive processes and USER time, and all of that drives load. When the chain moves together and IDLE never hits zero, your server is comfortable. When load climbs but channels do not, something other than call volume is eating the box.

For the full context on why these readings matter, see our guide to monitoring VICIdial server health and capacity. And once you can read the load graph, learn where the safe line sits for your hardware in safe CPU load thresholds for VICIdial.

If you would rather not stare at graphs to know your dialer is fine, our managed hosting reads them for you and sizes the box to your call volume. Compare the plans on the pricing page.