VICIfast
ENES
Guides & tutorials

Why every server in a VICIdial cluster must sit in one datacenter

Your dialers query the database many times a second. Put the machines in different cities and every answer arrives late, so live calls wait.

VICIfast Support
··4 min read
Servers in a data centre — illustration for: Why every server in a VICIdial cluster must sit in one datacenter
Photo by Sergei Starostin on Pexels

Should you put half your VICIdial servers in one city and half in another, so a bad day in one building can't stop the floor? It sounds like the careful choice. With VICIdial it works the other way round: the distance between the machines slows every call you take.

The dialers never stop asking the database

A cluster is one VICIdial system spread over several machines: a database and web server, two or more dialers running Asterisk (the phone engine that places and answers calls), and a standby database. If the parts are new to you, start with what each part of a cluster does.

The dialers don't work alone. They go to the database for the next leads to dial, for which agent is free, and to record every call, many times a second. On one outbound floor we measured, 166 Concurrent calls (calls up at the same moment) drove 2,230 database queries a second.

Each of those queries is a round trip. The dialer asks, waits for the answer, and only then acts on it.

What 20 milliseconds does to a floor

That wait is Latency, the time a request takes to get there and back. Inside one datacenter, on a private network, it is short enough that nobody notices. Between Germany and Finland it is over 20 ms.

Twenty thousandths of a second sounds harmless. It isn't, because the dialers pay it on every query they send, many times a second. A round trip that long is enough to stall VICIdial.

You feel it first at the hand-off. When a customer answers, VICIdial gives the call to a free agent through the database. While the dialer waits on those answers, the customer hears silence and the drop timer keeps counting. Slow it down enough and calls drop while your agents sit ready.

A split cluster can look healthy on every graph. In a cluster, a database problem shows up as a symptom on a dialer while CPU and memory look fine. Slow round trips never show on a processor chart; you see them as late hand-offs and dropped calls.

Agents and carriers can be far away. The servers can't.

Your agents already work from other cities, and your carriers sit somewhere else again. So why does distance matter here and not there? It comes down to who waits for whom.

Carriers connect to the dialers over a SIP trunk, the internet line that carries your calls. Agents load their screen from the database and web server and use a browser and a softphone for the phone side. That traffic is the calls themselves and a screen checking in once a second.

The traffic between the dialers and the database is a different kind. It is constant, it runs many times a second, and the dialer waits on each answer. That's why we put every machine in one datacenter on one private network, and only the outside world crosses the internet.

flowchart LR
  subgraph DC["One datacenter - private network"]
    D1[Dialer 1] -->|queries| DB[(Database + web)]
    D2[Dialer 2] -->|queries| DB
    DB -. replicates .-> SB[(Standby database)]
  end
  C[Your carriers] --> D1
  C --> D2
  A[Agents] -->|phone| D1
  A -->|phone| D2
  A -->|screen| DB

Where that one datacenter should be, relative to your carriers and agents, is its own question. We cover it in choosing a region for your dialer.

What about a second site for safety?

The worry behind a two-city layout is fair: you don't want one failure to stop the floor. A cluster handles that inside the building. With two or more dialers, a failed dialer drops the calls on it and the others keep the floor working.

The standby database replicates the main one as it changes, a live copy on the same private network. If the database machine fails, the standby is promoted to take over. That switch is called Failover.

If you need a floor that survives losing a whole region, that's a different design with its own trade-offs, not one cluster stretched across two cities. Our Enterprise work covers contracts and multi-region setups.

So the layout that protects you is one datacenter, one private network, and enough dialers that a failure stops only part of the floor. The VICIdial cluster guide walks through the rest of the layout.

If your floor still fits on one server, a single-server plan skips the question entirely. When you're ready to split, book a sizing call and see today's cluster prices.

Frequently asked

› Can I keep the standby database in another city?
Not in our layout. Every machine, standby included, sits in the same datacenter on one private network. Surviving the loss of a whole region is a separate Enterprise design.
› Do my agents need to be near the datacenter?
The one-site rule is for the servers, which talk to each other many times a second. Agents connect with a browser and a softphone, the same way they would to a single server.

About VICIfast

We run VICIdial servers for call centres, so you don’t have to. You get the dialer set up, secured and kept running — we handle the server, the updates and the backups. Bring your own phone carrier at no markup, or use VICIfast Voice, our own US calling and numbers, from $0.0085 a minute. From $49 a month per server.

Questions? Call +1 636 556 0022 and talk to someone who runs dialers.

Citing this article

VICIfast Engineering. “Why every server in a VICIdial cluster must sit in one datacenter”. VICIfast LLC, November 20, 2026. Retrieved from https://vicifast.com/blog/vicidial-cluster-one-datacenter

Have questions?

Related posts

You might be interested in

Comments

Comments are reviewed before they appear. We never publish your email.

No comments yet - be the first.