How to spot capacity problems before they hit your agents

Capacity problems rarely announce themselves. One day your VICIdial server is fine, the next your agents complain about choppy audio and the dialer feels slow. The good news is the warning signs show up in the reports days before agents feel them. Here is how to read the early signals and act while you still have room.

Watch peak load, not just average

The Server Performance Report summary gives you both average system load and peak system load. The average can look comfortable while the peak quietly creeps toward the ceiling. Agents do not feel the average. They feel the peak, the busiest moment of the shift when everyone is dialing at once. Track the peak number day over day and you will see the trend before any single day breaks.

Pair that with the average number of Asterisk channels in use, which reflects how many concurrent calls the box carried (see Concurrent calls). Asterisk is the telephony engine VICIdial runs on (see Asterisk), and every active call leg is a channel (see Channel). When average channels rise week over week, you are asking the box to do more, and load will follow.

Use the 30-day view to catch the trend

The Maximum System Stats report is built for exactly this. It draws a bar per day for the last 30 days and tracks total calls, most concurrent calls, and most concurrent agents. A single tall bar is noise. A staircase of bars climbing across the month is a capacity problem forming. When most concurrent agents keeps setting new highs, you are running closer to the edge every week.

How the signal turns into agent pain

stateDiagram-v2
  [*] --> Healthy
  Healthy --> Watch: peak load climbing on 30-day bars
  Watch --> Strained: average channels keep rising
  Strained --> AgentPain: IDLE near zero at peak
  AgentPain --> Healthy: add capacity or reduce dial level
  Watch --> Healthy: load settles back down

The path runs in stages. First the peak load creeps up on the 30-day bars. Then average channels rise. Then at the busiest moment IDLE processor time runs out, and only at that point do agents feel it as lag in the agent screen, a delay between calls (the screen connection lag is its own report under latency, see Latency).

There is a lever before you buy hardware. In predictive mode the dialer places several calls per available agent (see Predictive dialing), set by the dial level (see Dial level). Easing the dial level lowers channel and load pressure right away. It buys time, but if the 30-day trend keeps climbing, the real fix is a bigger box.

Build a simple weekly habit: note the week's peak load and the highest most-concurrent-agents bar. Two weeks of upward movement is your signal to plan capacity now, while you still have room, not during the shift when phones are ringing.

For the full monitoring picture, see our guide to monitoring VICIdial server health and capacity. And to put a clear line on when the trend means it is time to grow, read when a VICIdial server needs more capacity.

If you would rather have capacity handled for you, our managed VICIdial hosting watches these trends and resizes your server before agents ever notice. See how it scales on the pricing page.