Two taxi fleets: telephony, integrations and 1C under one SLA

Two taxi fleets are under our support in parallel: one operating in Moldova, the other abroad. Different countries, different telecom operators, different regulations — but the operational task is identical: pick up the customer call, hand the order to an available driver, and lose no contact along the way.

For a taxi service, telephony is not an "auxiliary service" — it is the service. A minute of PBX downtime means dozens of lost orders and customers who will phone the competition next time. Integrations with carriers and with the dispatch software aren't "to be done later" — they are part of the critical path from day one.

On top of telephony, each fleet has its own server infrastructure: business systems, 1C for driver settlements and reporting, file shares, backups. All of it requires not a one-off setup but ongoing engineering support with a clear SLA.

Each taxi fleet arrives with a similar set of tasks — with local nuances (carriers, regulations, integrations). We close them with a single engineering team:

• Stable 24/7 telephony — a taxi service runs round the clock, downtime is not on the menu.
• Simultaneous integration with multiple carriers — for fault tolerance and for coverage of different directions (city, mobile, international).
• GSM gateways and SIM banks — to receive calls from local mobile numbers and to avoid hitting outbound limits of a single carrier.
• PBX-to-dispatch integration — the customer card pops up on the dispatcher's screen with the call, the order is handed automatically to the closest available driver.
• Automated callback logic — if a call was missed or the customer hung up while queued, the system calls back at a configured time and logs the outcome.
• Server virtualisation — critical services (PBX, dispatch software, 1C, monitoring) on a fault-tolerant virtualised platform instead of "one service per box".
• 1C server support — configuration, scheduled maintenance, indexing, backups, remote access for accounting and management, technology log for diagnostics.

Telephony

The core is a VoIP PBX on an open-source stack (Asterisk / FreePBX) with several SIP trunks to different carriers and automatic failover when any of them goes down. In parallel, GSM gateways with SIM banks are connected — this delivers inbound calls from local mobile numbers and rotates outbound traffic across SIMs to stay within carrier limits and avoid blocks.

Integration with dispatch software

The PBX is wired to the dispatch software via AMI/AGI and REST webhooks. When a customer calls, the dispatcher sees a card with order history, rating and default addresses. A click in the dispatch UI fires an outbound call through the PBX immediately or queues the order for distribution to drivers.

Callback logic

Implemented at the dial-plan and PBX queue level. If a customer hung up before being picked up, or if the line was busy, the system places the number in the auto-callback queue. Based on CDR and configured triggers, the PBX calls back within the set interval, ties the result back to the call record, and on a successful connection raises the customer card on the next available dispatcher. Missed calls are visible to management in real time.

Server infrastructure

Virtualisation on a hypervisor (KVM/Proxmox or VMware ESXi, depending on the fleet) with multiple VMs on one physical node: PBX, dispatch software, 1C server, file server, monitoring, backup node. A standby hypervisor and replication of critical VMs deliver failover on hardware faults without long outages. Backup is centralised, with regular restore verification on a test bench.

1C support

The 1C server is sized for load, kept on a current release, and scheduled maintenance (integrity tests, reindexing, temp-table cleanup) runs on a schedule. Database backups are taken and verified by restore. Accounting and management get remote access via RDP/RemoteApp with role separation. When something stalls, the technology log helps diagnose heavy queries and locks.

Monitoring and unified SLA

Every component — PBX, GSM gateways, servers, 1C, backups — is wired into one monitoring stack, with alerts going to the on-duty rotation. Call metrics (missed calls, average wait, dispatcher occupancy) and infrastructure status sit on a single dashboard. Any critical-service outage is caught in the first seconds — not "when a customer notices".