Skip to content

Platform Reliability

Built to keep going when a piece of it stops

Two FreeSWITCH media nodes with agent failover at reconnect, load-balanced WebRTC, a SIP load balancer spreading calls across the nodes, readiness checks that look at real dependencies, reconcilers that clean up after a wobble, and nightly off-site backups.

14-day free trial · no credit card required

The system health view of the DialerBee operations center: the architecture from the agent browser and admin console through the reverse proxy to the eight backend services, each with its current response time, and an all-systems-operational state
Built by BroadNet, 22 years in telecom 11 languages, Arabic dialect-aware BYOC, your carriers Compliance-supporting controls

Quick answer

What is Platform Reliability in DialerBee? It is how the platform stays available and how it repairs its own state. Media runs on two FreeSWITCH media nodes, with agent WebSocket failover between them at reconnect, load-balanced WebRTC for browser-based agents, and a SIP load balancer distributing calls round-robin across the nodes. Readiness endpoints check a service's dependencies before it reports ready. Reconcilers correct orphaned calls, stuck agents and calls that were in flight across a restart, so nothing is quietly left behind. The database is backed up nightly and held off-site. The Enterprise plan carries a 99.95% uptime SLA. Node health, live calls and reconciler activity are visible in the operations center, so this is something you can watch rather than take on trust.

What you get

Everything in one workspace

Two FreeSWITCH media nodes

Media runs on two FreeSWITCH nodes rather than one. A single node is a single afternoon away from being your whole outage.

Agent WebSocket failover

The agent desktop holds a WebSocket to a media node. If that node goes away, the agent's client fails over to the other one at reconnect.

Load-balanced WebRTC

Browser-based agent audio is load balanced across the nodes, so the agents are spread rather than piled onto whichever node answered first.

Round-robin SIP

A SIP load balancer distributes calls round-robin across the media nodes, so carrier traffic does not depend on one node staying up.

Dependency-aware readiness

Readiness endpoints check the dependencies a service actually needs before reporting ready. A service that cannot reach its database should not be told to take traffic.

Orphaned-call reconciler

Calls that lost their controlling record are reconciled rather than left running. An orphaned call bills, holds a channel and helps nobody.

Stuck-agent reconciler

An agent stuck in a state they cannot leave is detected and put back into a real one, instead of sitting out the shift as a phantom.

In-flight call reconciler

Calls in flight during a restart or a hiccup are reconciled against reality afterwards, so the platform's view and the switch's view agree.

Nightly database backups

The database is backed up nightly. Not on request, not when someone remembers, on a schedule.

Off-site backups

Backups are held off-site, because a backup that shares a fate with the thing it protects is not a backup.

Uptime SLA on Enterprise

The Enterprise plan carries a 99.95% uptime SLA, written into the agreement rather than implied by a marketing page.

Visible from the operations center

Node health, live calls and reconciler activity are visible in the operations center, so reliability is something you can watch rather than assume.

Media

Two nodes, and a client that knows it

Media runs on two FreeSWITCH nodes. That alone is common enough. The part that decides whether an agent notices an outage is what the agent's browser does about it.

The agent desktop holds a WebSocket to a node, and when that node goes away the client fails over to the other node at reconnect. WebRTC is load balanced across the nodes, and a SIP load balancer sends carrier traffic round-robin across them, so neither the agents nor the carriers are pinned to one machine.

  • Two FreeSWITCH media nodes carrying media
  • Agent WebSocket failover between nodes at reconnect
  • Load-balanced WebRTC for browser-based agents
  • SIP load balancer distributing calls round-robin across the nodes
The infrastructure half of the DialerBee operations center's system health view: the two FreeSWITCH media nodes with their channels and registrations, the SIP trunk down to the carrier gateway, and the Postgres, Redis and object-storage dependencies beside them

Self-correction

The platform tidies up after itself

Contact centre platforms do not usually fail loudly. They fail by leaving things behind: a call that is still up with nothing controlling it, an agent stuck in a state they cannot leave, a call that was in flight when a service restarted.

Reconcilers exist for all three. They compare what the platform believes with what is actually happening and correct the difference. Alongside them, readiness endpoints check a service's real dependencies before it reports ready, so nothing is handed traffic it cannot serve.

  • Reconciler for orphaned calls
  • Reconciler for stuck agents
  • Reconciler for in-flight calls across restarts
  • Readiness endpoints that check dependencies before reporting ready
The campaign management screen in the DialerBee admin console for the Northwind Demo tenant: seven campaigns in one table, each row showing its status of draft, stopped, completed or paused, its dialing mode of progressive or preview, its gateway, its queue, and view, start, edit and delete actions, with a search box and status and mode filters above the table

Backups and commitment

Nightly, off-site, and in writing

The database is backed up nightly and the backups are held off-site. Both halves matter. A nightly backup sitting on the same infrastructure as the platform is a copy, not a recovery plan.

On top of that the Enterprise plan carries a 99.95% uptime SLA. It is in the agreement, which is where an availability commitment belongs, rather than on a slide.

  • Nightly database backups
  • Backups held off-site
  • 99.95% uptime SLA on the Enterprise plan
  • Node health, live calls and reconciler activity visible in the operations center
The recordings screen in the DialerBee admin console for the Northwind Demo tenant: 459 recordings with three on legal hold and a 30-day retention note, summary cards for recording count, total duration, storage used and agents, and a table of calls with date and time, direction, agent, customer phone number redacted, duration, status, legal-hold state and play and download actions

How it compares

A single-box deploymentDialerBee platform reliability
MediaOne nodeTwo FreeSWITCH media nodes
When a node goesAgents reload and hopeAgent WebSocket failover between nodes at reconnect
Carrier trafficPointed at one addressSIP load balancer, round-robin across the nodes
Agent audioWhatever answered firstLoad-balanced WebRTC across the nodes
Health checksIs the process aliveReadiness endpoints that check the service's dependencies
Leftovers after a wobbleSomeone finds them laterReconcilers for orphaned calls, stuck agents and in-flight calls
BackupsWhen someone remembersNightly database backups, held off-site
CommitmentBest effort99.95% uptime SLA on the Enterprise plan

Frequently asked questions

What is Platform Reliability in DialerBee?+

It is how the platform stays up and how it corrects itself when something slips. Media runs on two FreeSWITCH media nodes with agent WebSocket failover between them at reconnect, load-balanced WebRTC and a SIP load balancer distributing calls round-robin across the nodes. Readiness endpoints check dependencies, reconcilers clean up orphaned calls, stuck agents and in-flight calls, and the database is backed up nightly and held off-site.

What happens to my agents if a media node fails?+

The agent desktop holds a WebSocket to a media node. If that node becomes unavailable, the agent's client fails over to the other node at reconnect, so the agent comes back on the surviving node rather than waiting for the failed one.

How is call traffic spread across the nodes?+

A SIP load balancer distributes calls round-robin across the media nodes, and browser-based agent audio is load balanced over WebRTC. Neither the carriers nor the agents are tied to a single node.

What are the reconcilers for?+

They correct the state that gets left behind when something goes wrong. One reconciles orphaned calls, calls still up with nothing controlling them. One reconciles stuck agents, agents left in a state they cannot leave. One reconciles calls that were in flight across a restart. Each one compares what the platform believes with what is actually happening.

What does dependency-aware readiness mean?+

A readiness endpoint checks the dependencies a service actually needs before reporting that it is ready. A service that cannot reach what it depends on is not sent traffic, which is the difference between a health check and a useful health check.

How are backups handled?+

The database is backed up nightly and the backups are held off-site. Off-site is the part that matters, because a backup stored alongside the platform shares whatever happens to the platform.

Is there an uptime SLA?+

Yes, a 99.95% uptime SLA on the Enterprise plan. It sits in the agreement rather than on a marketing page.

Can I see any of this myself?+

Yes. The operations center shows node health, live calls and reconciler activity, along with alerts you can configure for a NOC. Reliability you can watch is more useful than reliability you are told about.

Ask us how DialerBee behaves when a node goes away

A 30-minute walkthrough on a demo tenant, in English or Arabic.