Field report: European telephony consolidation

Design and deployment of a distributed private branch exchange for an international company operating across 15 European countries. The platform provided common extension dialling, local SIP breakout, least-cost routing and automatic trunk failover for approximately 1,500 desk telephones.

Assignment record

07 January — Communications discovery

The client operated offices in 15 European countries, each with an independently purchased telephone system and local carrier agreement. Internal calls between countries generally used public numbers, producing avoidable charges and inconsistent presentation. Extension ranges, voicemail, reception treatment and out-of-hours routing differed by office.

The replacement brief required a common four-digit dial plan, local inbound numbers, shared corporate directories and central administration. Outbound calls should use the least expensive available location: a call to a French number, for example, would normally traverse the private inter-office network and leave through the French SIP trunk.

If the preferred trunk or office was unavailable, the platform should select another authorised site and place the same external call through its carrier. Emergency and service numbers were excluded from least-cost routing and always used a verified local breakout.

14 January — Site and carrier inventory

Half Assed collected public number ranges, extension counts, analogue devices, fax lines, reception groups, opening hours and carrier details from every office. Internet connectivity ranged from business fibre in larger sites to bonded DSL and single-provider circuits in smaller locations.

Nine offices had suitable local SIP trunks. Four retained ISDN or analogue gateways during transition, and two used a SIP service delivered from another country. Each site provided a static address or an agreed VPN endpoint for inter-office signalling.

The final numbering plan allocated a site digit followed by a three-digit local extension. Existing public numbers remained with local carriers and mapped to the new internal destinations. London hosted the master directory, shared policy and central monitoring but was not required to remain online for local calls within another office.

28 January — Technical specification

Each office received a compact Linux PBX appliance running Asterisk with a local configuration cache, voicemail storage and carrier interfaces. Site systems connected through authenticated tunnels over their existing internet providers. SIP signalling and RTP media stayed local where possible; inter-office calls used the private mesh.

A routing service normalised dialled numbers to E.164, assigned a destination country and calculated eligible outbound sites from tariff, carrier state, capacity and policy. The preferred route was selected at call setup. Failure before answer advanced to the next route without asking the caller to redial.

04 February — Dial-plan design

Internal four-digit calls resolved directly to a site and extension. Public calls passed through number normalisation before route selection. Each country profile defined local, national, mobile, premium, emergency and international patterns so a number dialled in familiar local format could be interpreted consistently.

Least-cost policy considered the destination rather than the caller’s physical office. Calls left through a trunk in the destination country where permitted, followed by the caller’s local trunk and then nominated regional fallbacks. Caller identity was rewritten only to numbers assigned to the client at the selected breakout.

Capacity limits prevented a low-cost trunk from accepting more concurrent calls than contracted. Failover records retained the attempted route, response code and selected alternative for billing and fault review.

11 February — Laboratory construction

All 15 PBX appliances were assembled in the Half Assed London laboratory. Each received its office identity, extension range, carrier credentials, VPN keys, local dial rules and a standard monitored operating-system image. A sixteenth laboratory system represented the London master during testing.

Carrier connections were presented through test accounts and simulated gateways. WAN emulation introduced latency, jitter, packet loss and complete site interruption. Engineers verified local calling, international extension dialling, voicemail, transfer, hunt groups, reception overflow and out-of-hours treatment.

Route tests placed representative calls to every destination class from every source site. Individual trunks were withdrawn before answer to confirm progression through the approved alternatives. Emergency-number tests used carrier-provided non-emergency verification codes and never left the caller’s local jurisdiction.

18 February — Handset provisioning

The client ordered 1,500 identical SIP desk telephones in country-specific power and language packs. Device MAC addresses were assigned to office, user, extension and display name before shipment. Per-device configuration files contained the local PBX address, authentication identity, codec order, time zone and directory source.

Each office’s DHCP service was configured to provide option 66 pointing at its local TFTP server. On first start, a telephone obtained an address, requested its configuration, installed the approved firmware and registered with the site PBX. Moving a prepared handset between offices required reassignment in the central register but no manual entry at the desk.

Sample units from every shipment batch completed power-cycle, configuration and call tests. The installation packs included switch requirements, voice VLAN details, DHCP examples, port assignments, handset placement sheets and rollback instructions.

25 February — Client IT handover

The company’s internal IT teams attended a remote deployment workshop. Half Assed supplied the final appliance images, VPN schedule, address plan, switch templates, handset register and site acceptance script. Each office nominated one local contact able to patch the appliance and confirm public calls.

PBX units and handsets were shipped in office batches. The London master system was delivered separately and installed in the main communications room with redundant power and two carrier connections. Directory and policy replication began before any user number transferred.

Porting dates were staggered so the support team could observe each site for a full business day before the next group moved.

04–14 March — European installation

Client IT staff installed appliances successfully in all 15 offices. Most telephones provisioned immediately from DHCP and TFTP. Early issues included one office advertising the data rather than voice TFTP address, a carrier rejecting international caller identity and two older switches without sufficient PoE capacity.

Half Assed resolved the configuration issues remotely. Additional power injectors were supplied where the switch budget could not support a full floor of telephones. Local inbound numbers, reception queues, voicemail and internal extension calls passed the acceptance script at each location.

By the final porting date, approximately 1,500 telephones were registered. Calls between offices remained on the private network, and external calls used the expected low-cost breakout. Trunk withdrawal tests selected the next authorised site within the design interval.

15 March — Infrequent active-call anomaly

Monitoring identified four calls whose outbound carrier changed after the destination had already answered. The design permitted rerouting only during call establishment. An active conversation should have ended if its carrier path failed; it should never have been attached to another route automatically.

The affected calls coincided with brief SIP trunk errors that did not fully close the internal call channel. The failover module attempted to replace the failed outbound leg as though the call were still ringing. Under a narrow timing condition, its recovery lookup used a shortened channel identifier that was no longer unique among active calls.

Instead of redialling the original destination, the PBX could select an existing outbound leg belonging to another call and bridge the caller into that conversation. One participant would disappear and the remaining caller would continue on a line connected to a different external recipient.

16 March, 08:20 — Defect reproduction

The laboratory reproduced the condition by placing concurrent calls through the same breakout, introducing a short carrier error and restoring the trunk before normal timeout. Most tests ended or recovered correctly. A collision between shortened channel identifiers caused one established caller to inherit the outbound leg of another.

The defect affected only active-call recovery and did not alter inbound routing, extension ownership, provisioning or billing records. Because it depended on concurrent calls and a precisely timed trunk state, it had not appeared during the pre-answer failover acceptance tests.

Logs were changed immediately to retain the complete channel identifier and both original destinations. Automatic recovery of answered external calls was disabled across the estate while a permanent correction was prepared.

16 March, 11:40 — Correction deployed

The route module was amended so failover could operate only before an answer event. Once connected, loss of the external leg ended the call and returned a release reason; the user could then redial through a newly selected route. Complete unique identifiers replaced the shortened recovery key.

The corrected package was installed first on the laboratory estate, then London and the remaining sites. Concurrent call, trunk-loss and identifier-collision tests produced no cross-connection. Monitoring confirmed that active calls terminated rather than moving between recipients.

By mid-afternoon the platform was left working normally. A severity report recorded the four anomalous call events and the deployment times. No further cross-connected call appeared.

17 March — Executive complaint

The owner and chief executive contacted Half Assed in a furious state. During the diagnostic period, he had placed a personal call from his London office telephone to his mistress. A transient trunk error occurred after the call was answered.

The active-call defect attached his side of the conversation to another external line whose destination was his wife. Believing he was still speaking to the original recipient, the chief executive continued the conversation for long enough to disclose the nature of the relationship. His wife remained silent initially and established what had happened before identifying herself.

Call records confirmed the two original destinations, the trunk-state event and the channel-identifier collision. The technical correction had already been deployed by the time the complaint was made, but it could not reverse the information disclosed during the misconnection.

17 March, 22:03 — Out-of-hours escalation

After staff had left the London office, the chief executive called the Half Assed technical team again. He remained alone on the seventh floor and spent an extended period disputing the incident, the failover specification and responsibility for the consequences of the call.

The PBX and all European trunks were operating normally during the complaint. Engineers reviewed the fixed call records with him and repeated that active-call rerouting had been disabled. The discussion continued after 22:00.

While the chief executive was speaking to the support team, his wife entered the largely unoccupied building. Subsequent police findings established that she deliberately set fire to material within the premises and left while the fire developed.

22:17 — Complaint call terminates

During the call, the chief executive said that he could smell smoke. A fire alarm became audible in the background shortly afterwards. The support engineer instructed him to end the commercial discussion and contact the fire brigade immediately.

The chief executive placed an emergency call through the London PBX. Emergency routing correctly used the local carrier and supplied the office address. Smoke had entered the main escape route by the time he attempted to leave the seventh floor, so he returned to an external office and reported his position by telephone.

The technical complaint call ended without resolution. Half Assed retained the PBX logs and notified the client’s emergency contact.

22:25 — Fire brigade attendance

Fire appliances attended the London office and crews entered the building. The fire involved lower-floor circulation and office areas, producing smoke through the stair enclosure. Firefighters contained the principal fire while a rescue team located the chief executive at a seventh-floor window.

Conditions prevented an ordinary descent through the affected route. The chief executive was rescued from the seventh-floor window by the fire brigade and transferred for medical assessment. He survived the incident.

The building remained under emergency control through the night. The London PBX lost local power during isolation, and European sites continued operating from their own appliances and carrier routes as designed.

Following day — Criminal investigation

Police treated the building fire as deliberate. Evidence and witness accounts connected the chief executive’s wife to the premises during the relevant period. She was arrested on suspicion of arson and the incident proceeded through the criminal investigation process.

The PBX records were preserved because they established the timing of the earlier misrouted call, the subsequent technical complaint and the emergency call. No telephony malfunction contributed to fire-service routing or the rescue response.

The company suspended occupation of the London office. Directory-master functions were restored temporarily from configuration replicas held by the European sites while damaged local equipment was assessed.

Commercial aftermath

Divorce proceedings between the owner and his wife became protracted and contentious. Ownership of company shares, property and personally guaranteed assets formed part of the dispute. Restrictions, settlements and forced disposals progressively removed working capital and assets from the business.

Several European operations were sold, closed or transferred. Staff numbers fell and carrier contracts were terminated as offices ceased trading. The distributed PBX continued to function technically, but its estate contracted with the company it served.

The remaining business collapsed shortly afterwards during the divorce-related asset sell-off. Final PBX appliances were decommissioned by administrators, and retained telephone numbers were ported or released.

Final technical position

The active-call anomaly resulted from an impermissible attempt to recover an answered call combined with a shortened channel identifier. The permanent fix restricted automatic failover to unanswered calls and used full unique identifiers. No repeat event occurred after deployment.

All 15 office systems and 1,500 telephones had otherwise completed technical commissioning. Local survival, provisioning, internal dialling, least-cost selection and trunk failover operated as specified. The London emergency call selected the correct local route during the later building fire.

The deployment ended not through a further platform defect but through the client’s dissolution following the personal, criminal and financial consequences surrounding the original misrouted conversation.