What does VICIdial need for more than 100 agents?
From about 100 agents or 500 concurrent calls, VICIdial is usually split into a cluster, and from 200 predictive agents no single server carries the floor. Up to 150 predictive agents still fit one server - Hyper, with 48 dedicated cores, at about 59% at the busiest minute - and the calculator shows both options for that size below. VICIfast runs either as managed VICIdial: the same VICIdial admin and agent screen, with every server patched, firewalled and monitored for you.
| How the floor dials | Calls at peak | Smallest server that fits | Peak CPU |
|---|---|---|---|
| Outbound, predictive 3 lines per waiting agent | 450 | Hyper · 48 dedicated cores · $3,249/mo | 59% |
| Inbound every agent on a call, 10% more waiting | 165 | Turbo · 32 dedicated cores · $1,899/mo | 47% |
| Manual dialing one line per agent | 150 | Turbo · 32 dedicated cores · $1,899/mo | 45% |
- Bandwidth at peak
- 53 Mbps
- Predictive outbound, both legs
- Carrier channels
- 540
- Concurrent calls to ask your carrier for
- Recordings per day
- 45.4 GB
- 8 dialing hours
- Days of recordings on disk
- 20
- On Hyper, before offloading
| Role | Servers | Each |
|---|---|---|
| Database | 1 | Velocity · 16 dedicated cores · $899/mo |
| Database replica | 1 | Velocity · 16 dedicated cores · $899/mo |
| Web and agent phones | 5 | Surge · 8 dedicated cores · $449/mo · 30 agents |
| Dialer (Asterisk) | 2 | Pro · 8 shared cores · $399/mo · 225 calls |
The first table is one server per dialing mode for 150 agents. The second is the cluster layout for the predictive row: what each role runs on, and how many agents or calls each server carries. The agent counts on pricing are suggestions for a typical mix; at this size, how many lines each agent dials matters more than the headcount.
How big is a VICIdial cluster for 150, 200 or 300 agents?
A VICIdial cluster for 150 to 300 predictive agents is a database and replica plus five to nine web servers and two to four dialers, growing with agents and live calls. Agents decide how many web servers you need; live calls decide how many dialers. At three lines per agent:
| Predictive agents | Live calls at peak | One server? | Cluster layout |
|---|---|---|---|
| 150 | 450 | Hyper, about 59% CPU | database + replica, 5 web, 2 dialers |
| 200 | 600 | No single server fits | database + replica, 6 web, 3 dialers |
| 300 | 900 | No single server fits | database + replica, 9 web, 4 dialers |
The rules behind it: one web server for about every 35 agents, since each also runs those agents' phones, and one Asterisk-only dialer for about every 250 live calls on 8 cores - never fewer than two, so losing one does not stop every call. Inbound floors stay on one server much longer: 200 inbound agents fit Turbo and 300 fit Hyper. The VICIdial server sizing guide explains the single-server side of the numbers.
Why do the database and reports matter at this size?
Because every server in a VICIdial cluster works from one MySQL database - agent states, the lead hopper, call logs - so the database is the one part you cannot split, and anything that slows it slows every agent on the floor. That is why the layout starts with the database, why it comes with an identical replica, and why reports read the replica instead of the live database. A supervisor pulling a month of call detail at 11 in the morning should not cost the floor dial time. Agents and supervisors still see one floor: the same admin, the same agent screen and the same real-time report, whichever web server they log in to. VICIdial reports explained walks through the reports themselves, and reporting shows what VICIfast adds on top.
What does enterprise delivery include?
Clusters for large floors come through VICIfast Enterprise, delivered in under 24 hours. What comes with it:
- Multi-region VICIdial clusters on dedicated hardware. 23-node clusters have been delivered in 14 hours.
- A named technical contact (TAM) for your account.
- A 99.99% uptime SLA, or a custom SLA, with a 15-minute response on severity-1 issues.
When doesn't a large floor need a cluster?
When it is inbound or manual: a single Hyper server carries 300 inbound agents. A predictive floor of 150 at one site that can accept one server as a single point of failure can also stay on Hyper - but Hyper is the largest plan, so the next step from there is a cluster. Scaling VICIdial vertically covers the trade-off of one bigger server against several smaller ones. One caveat on the numbers: the one-server rows come from a model we check against real shifts on our servers, while the cluster rows are our build rules, which we check against your floor before anything is built.
Where to start
Put your agents, lines per agent and inbound calls into the capacity calculator to see where one server stops and what the cluster looks like. Then talk to us or go straight to enterprise with the layout - a person reads every message.