What makes VICIdial a high volume dialer?
VICIdial becomes a high volume dialer when the server, the carrier and the database are all sized for the number of live calls, not the number of agents.
VICIdial is free, open-source call center software built on Asterisk. The software is not the limit: CPU, carrier channels and database load are, and each of them grows with live calls.
VICIfast is managed VICIdial hosting. You get your own VICIdial server in under 40 seconds, with patching, firewall, SSL and 24/7 monitoring handled, and you keep full root access. Pricing is per server, not per agent, so a high call volume does not mean a per-seat bill.
What limits call volume on VICIdial?
Four things limit volume on VICIdial: CPU per live call, carrier channels, database load from reports, and the latency your agents feel.
- CPU per live call. Every live call is an Asterisk channel, and recording it adds work. The VICIfast capacity model records every call by default, so its numbers include that work.
- Carrier channels. Your SIP carrier caps how many calls you can have up at once. A predictive floor needs far more channels than it has agents.
- Database load. Leads, live agent state and call logs all sit in one database. A heavy report run in the busiest hour competes with the dialer for the same disk and CPU.
- Agent latency. When the server runs short, agents feel it first: lag between calls, or being logged out in the busy hour. Latency is often the earliest sign you are out of capacity.
How many live calls does a high volume floor carry?
A predictive agent dialing three lines carries about three live calls, so 100 predictive agents is about 300 live calls - nearly three times what 100 inbound agents need.
The agent counts on our plans are suggestions for a typical mix, so a high call volume VICIdial floor dialing predictive needs more cores than the plan name suggests. The capacity calculator's results, with every call recorded:
| Agents | Dialing | Live calls | What it runs on |
|---|---|---|---|
| 50 | Inbound | 56 | Pro |
| 50 | Predictive | 150 | Ultimate, or Velocity on dedicated cores |
| 60 | Predictive | 180 | Velocity |
| 70 to 80 | Predictive | 210 to 240 | Turbo |
| 100 | Inbound | 111 | Velocity |
| 100 | Predictive | 300 | Turbo (58% peak CPU), or a cluster: database + replica, 3 web servers, 2 dialers |
| 150 | Predictive | 450 | Hyper (59% peak CPU), or a cluster: database + replica, 5 web servers, 2 dialers |
| 200 | Predictive | 600 | Cluster: database + replica, 6 web servers, 3 dialers |
| 300 | Inbound | 330 | Hyper |
| 300 | Predictive | 900 | Cluster: database + replica, 9 web servers, 4 dialers |
Put your own agent count, dialing mode and recording choice into the agent capacity calculator to see the plan and peak CPU for your floor.
When does one VICIdial server stop being enough?
For a predictive floor, one server stops being enough somewhere between 150 and 200 agents, and the calculator suggests a cluster earlier, from about 100 agents or 500 concurrent calls.
Hyper, the largest dedicated plan, is suggested for about 350 agents, but that is a typical mix of calling. At three lines per agent, the ceiling arrives much sooner. Past it, VICIdial at scale runs as a cluster with three kinds of server:
- A database and an identical replica. Reports read the replica, so a heavy report no longer slows the dialer.
- Web servers that also run the agents' phones, about one for every 35 agents.
- Asterisk-only dialers, each carrying about 250 live calls on 8 cores, with at least two.
VICIfast enterprise clusters run on dedicated hardware and are delivered in under 24 hours - 23-node clusters have gone live in 14 hours - with a named technical contact, a 99.99% uptime SLA and a 15-minute response on sev-1 issues. You get a fixed-price quote within 48 hours.
How do you keep the drop rate under 3%?
Keep abandoned calls under 3% by tuning the dial level rather than the agent count, and by watching the live abandon rate every hour.
US rules cap abandonment at 3% per campaign over a rolling 30 days. One bad day will not sink a 30-day average, but a run of them will, so most floors run with headroom:
- Set the campaign's drop window to 2 seconds. New campaigns start at 5.
- Start the dial level at 1.0, and raise it in small steps only while agents sit idle more than 30% of the time and drops stay under 1%.
- If drops climb past 2%, lower the dial level by 0.2 and watch. Most teams settle between 1.5 and 2.5.
- Use an adaptive dial method if your connect rate changes through the day, and switch small lists of under about 5,000 leads to progressive or manual dialing.
Sizing at three lines per agent leaves room above that 1.5 to 2.5, so a good hour does not find the server short. The predictive dialer guide works through the maths, and our predictive dialing best practices cover the settings.
How many carrier channels and how much storage do you need?
Ask your SIP carrier for more channels than your peak live calls, and plan for about 0.3 GB of recordings per agent per 8-hour day.
| Predictive agents | Bandwidth | Channels to ask for | Recordings per 8-hour day |
|---|---|---|---|
| 50 | 18 Mbps | 180 | 15.1 GB |
| 60 | 21 Mbps | 220 | 18.1 GB |
| 70 | 25 Mbps | 260 | 21.2 GB |
| 80 | 28 Mbps | 290 | 24.2 GB |
| 100 | 35 Mbps | 360 | 30.2 GB |
You can use your own SIP carrier on every plan, or VICIfast Voice. For how channels, agents and servers relate, see sizing a VICIdial server by agents and channels.
Where VICIfast is not the right fit
If compliance rules require the dialer on hardware you physically control, a hosted cluster is the wrong shape. We can still install and manage VICIdial on a server you already own, for a flat $29 a month.
Where to start
Put your agent count and dialing mode into the capacity calculator, then read the VICIdial server sizing guide for the signs a server is too small. Under about 100 agents, start a free trial and choose the plan the calculator picks; if you grow, you resize to the next plan from the dashboard with about three minutes of downtime. For 100 agents or more, or 500 concurrent calls, tell us your numbers on the enterprise page and we will quote the cluster.