There is a specific kind of ViciDial failure that only appears when things are going well. The floor is full, pacing is up, calls are flowing, and then calls start failing, not for one agent but across the room, and the harder you push the worse it gets. That is a capacity ceiling, and it is one of the most misdiagnosed problems on an outbound floor. Here is how to find it and fix it.
Concurrency is the number that matters
Outbound floors tend to think in calls per hour, but the number that breaks things is concurrent calls, how many are in progress at the same instant. That figure is driven by your agent count, your dial ratio, and how long calls last, and it spikes well above your average whenever pacing climbs or calls run long.
Every part of your path has a ceiling on concurrent calls. Your server can only process so many audio streams. Your trunk is provisioned for only so many simultaneous calls. When your concurrency rises past one of those ceilings, calls start failing, and because concurrency is bursty, the failures look intermittent even though the cause is a hard limit. The first troubleshooting move is to watch concurrency, not call volume, and see whether failures track it. That same reasoning runs through our broader ViciDial connectivity field guide.
Server ceiling or trunk ceiling
Once you know failures track concurrency, the next question is which ceiling you are hitting, because the fix is completely different.
A server ceiling looks like the box struggling. CPU climbs, load average rises, audio processing falls behind, and the machine itself is clearly working hard. The fix there is your infrastructure, more resources or better tuning.
A trunk ceiling looks different. Your server has headroom to spare, but calls are refused or fail at the carrier as concurrency rises. The box is fine; the pipe is full. That is a carrier capacity limit, and no amount of server tuning fixes it, because the constraint is how many simultaneous calls your trunk is allowed to carry. Telling these apart is the single most useful diagnostic step, and it decides whether you are buying hardware or having a carrier conversation.
What to ask a carrier about load
If the ceiling is the trunk, the real question is whether your carrier was built for the load an outbound floor puts on it. A dialer does not make calls like an office phone system; it makes bursts of simultaneous calls that stress capacity in a way steady traffic never does. Plenty of carriers are fine for even office traffic and fall over under a dialer's burst.
So the questions to ask are specific:
- Does the trunk hold at my peak concurrency, with headroom? Not my average, my spike.
- What happens to completion as I push volume? Capacity that technically accepts calls but completes fewer under load is still a ceiling.
- Can I raise capacity quickly as I grow? A floor that scales needs a carrier that expects it, not a slow process every time.
These are part of the same evaluation covered in our checklist for choosing a termination carrier for a dialer and our carrier guide for ViciDial and Asterisk call centers. A carrier that connects by IP, holds capacity under burst, and scales with you is what keeps the room running when pacing is up. Related failures like post dial delay often show up alongside capacity strain, so it is worth watching for them together.
The failures that only appear when the floor is busy are almost always a ceiling, not a glitch. Measure concurrency, find out whether the limit is your box or your trunk, and if it is the trunk, get on a carrier that was built for the way a dialer actually loads a pipe.