NEXIATwo different products with the same name
"Call center software" means two things with nothing in common:
- The contact platform — dialer, agent desktop, queues, scripts, recording, CRM integration. Vicidial, GoAutoDial and the commercial suites live here. NEXIA Switch is not this and does not try to be.
- The switch that gives that traffic a way out — the one that takes the dialer's calls, picks which carrier they leave through, enforces limits and bills them. That is a Class 4 softswitch, and that is what NEXIA is.
The confusion matters because whoever buys the first looking for the second ends up with a platform that serves agents but has no idea what the minute cost.
Why dialer traffic is different
A switch sized for normal conversational traffic behaves differently when a predictive dialer points at it. Four differences, all awkward:
| Conversational traffic | Dialer traffic | |
|---|---|---|
| Calls per second (CPS) | Low and steady | Bursts when the campaign starts |
| ASR (answer rate) | High | Low: most calls go unanswered |
| ACD (average duration) | Minutes | Seconds, or zero |
| Attempts per billed minute | Close to 1 | Dozens |
| What costs you | The minutes | The signalling, which nobody bills |
That last row is the entire business: with dialer traffic, most of the switch's work goes into calls that are never charged for. If each failed attempt costs a system process and a database query, the server falls over at a ridiculous level of revenue.
What the switch has to solve
1. Admission that survives bursts without dropping or blocking the customer
When a campaign starts, the dialer fires everything at once. The switch has two bad ways to react: fall over, or treat the burst as an attack and block the customer. The right one is a CPS limit per customer, enforced at the edge, that queues or rejects the excess with a code the dialer understands — without touching anyone else's traffic.
2. The right rejection code, which is not a detail
This is the technical difference that moves the most money and gets the least attention. When you reject a call, the SIP code you return tells the customer's dialer what to do next:
- A
503means "temporary problem": the dialer retries, and sends you the same number again right away. - A
486or a404means "this number is a dead end": the dialer closes it out and moves on.
If your switch answers 503 to "no balance" or "no route", you have just
asked the dialer to flood you with retries of something that will never work. A serious
switch lets you
choose the code per
rejection cause.
3. One attempt per number, not one per gateway
The classic pattern: a dead number comes in, route 1 rejects it, LCR sends it to route 2, route 2 rejects it, on to route 3… The same dead number walks every carrier you have. Multiplied by an old contact list, that is millions of attempts that bill nothing, wreck your ASR with every carrier, and downgrade the route quality they give you. The defence is a guard of one attempt per number, not per gateway.
4. Channel caps enforced at admission
A customer with a misconfigured dialer can open a thousand channels by accident. The cap has to apply before any resource is reserved, and at every level: per customer, per connection point and per outbound route.
5. Margin-based LCR, because the margin here is thin
Termination for call centers sells at fractions of a cent. An LCR that sorts only by cost can take a route that, after the billable minimum and the connection charge, loses money. Sorting by guaranteed minimum margin is what makes the volume worth having.
6. ASR and ACD visible per customer and per route, today
With dialer traffic, a route's quality degrades in hours, not weeks. You need ASR by carrier and by destination in the moment, not in Monday's report.
Who this is for
- The carrier selling termination to call centers. Buys minutes from several carriers, sells them to several BPOs, and needs per-call margin and per-customer caps so they do not end up financing anyone by accident.
- The call center group with its own carriers. Stops depending on a single supplier, routes by cost and quality, and finally sees what the minute really costs per campaign.
- The reseller sitting in the middle today with a spreadsheet and an account on someone else's switch, who wants their own platform.
What NEXIA Switch does not do, so there are no surprises: no agent desktop, no predictive dialer, no queues, no IVR, no call recording for quality assurance, no CRM integration. All of that lives in the call center's contact platform, on the other side of the SIP link. NEXIA is the switch, not the contact center.
How NEXIA Switch handles it
All six points above ship by default: Kamailio admission at the edge with per- customer CPS limits, rejection codes configurable per cause, a repeat-dial guard keyed on the number, channel caps at five levels, margin-based LCR and CDRs with cost, sell and margin on the same row. The routing core is compiled and resolves in memory — it does not fork a process per call, which is exactly what makes failed-attempt traffic viable.
It is free and permanent up to 100 concurrent calls, with the complete platform, and installs with one command on Ubuntu 24.04.