VoIP network
A carrier-grade voice fabric engineered for cross-border business communication.
The Madrid Integration Services network is built on standards-based SIP, quality-weighted routing, and redundant interconnects across six regions — operated transparently and supported by engineers.
Infrastructure overview
The components, in plain terms.
The network is a small number of well-understood pieces working together. We don't believe in adding complexity that has to be defended later.
Session Border Controllers
SIP normalization, topology hiding, and media anchoring across all customer edges.
Core routing engines
Quality-aware least-cost routing with geographic and time-of-day policy overlays.
Tier-1 interconnects
Direct relationships with carriers across the USA, UK, EU, APAC, and a dedicated China gateway.
Media handling
G.711, G.722, Opus, and codec negotiation tuned for clarity over compression ratio.
Architecture at a glance
Customer edge to carrier — four well-defined stages.
The diagram shows how voice traffic moves through the platform. Each stage is independently observable, independently scalable, and independently recoverable. That separation is what makes complex incidents tractable.
- Customer-side endpoints normalised at the SBC
- Routing decisions made on quality and policy signals
- Carrier interconnects selected per call, per leg
- Continuous monitoring at every transit point
Routing architecture
How calls are placed on the network.
Routing logic is the most visible part of any VoIP service. It deserves to be explained, not hidden behind black-box marketing.
Quality-weighted least-cost routing
Routes are ranked by composite signals — ASR, MOS, post-dial delay — before cost is considered. Cost is a tiebreaker, not the primary axis.
Geographic and time-of-day policy
Inbound calls can be routed based on origin country, dialed DID, or business hours. Outbound calls can be shaped to honour carrier-specific quality windows.
Automatic re-attempt logic
When a leg fails or quality degrades mid-call setup, alternate paths are evaluated and re-attempted within signalling timeouts.
Carrier-aware substitution
If a carrier shows quality degradation against established baselines, traffic is quietly shifted to alternative routes while we engage the underlying provider.
Global DID & numbering
Local numbers with consistent operational behaviour.
Numbering plans differ meaningfully by country. We treat each market on its own terms while maintaining a single provisioning experience for customers.
United States
Local, toll-free, and geographic DIDs with FCC-aligned origination and STIR/SHAKEN-aware caller identity practices.
United Kingdom
01, 02, 03, and 080x ranges with awareness of Ofcom numbering rules and end-user verification expectations.
Canada
Geographic numbers across major provinces with interoperability into North American termination paths.
Australia
02, 03, 07, 08 geographic numbers and 1800 ranges, with carrier interconnects across major metros.
Germany
Geographic numbers aligned with BNetzA expectations on number ownership and registered usage.
China
Inbound and outbound voice paths through a dedicated gateway architecture designed for cross-border continuity.
SIP trunking & call flow
Standards-based signalling, predictable behaviour.
A SIP trunk is only as good as its behaviour under unusual conditions. These are the details we hold ourselves to across every customer integration.
RFC 3261 compliant signalling
Standards-based SIP with consistent INVITE / 18x / 200 OK handling across all entry points.
TLS / SRTP support
Encryption available end-to-edge where customer environments and carrier paths permit it.
Codec strategy
Default to G.711 for clarity on stable links, Opus for adaptive networks, with explicit negotiation rules.
NAT traversal
STUN, ICE, and SBC-side media anchoring to handle customer-side NAT realities without manual workarounds.
DTMF handling
RFC 2833 / 4733 in-band signalling with consistent inter-digit timing for IVR compatibility.
Call detail accuracy
CDRs with reliable post-dial delay, answer time, and disconnect cause coding for downstream analytics.
Redundancy & failover
No single point of failure earns the right to one.
Failure modes are mapped before they happen.
Each layer of the platform has documented failure scenarios, expected behaviour during the failure, and the path back to normal operation. We rehearse these regularly — incidents are not an opportunity to learn the runbook.
Signalling redundancy
Multiple SIP entry points across geographically distinct PoPs. Loss of any single PoP does not interrupt new call setup.
Media redundancy
Anchored media with re-anchor capability if a media server experiences a fault, transparent to most user agents.
Interconnect redundancy
Each region has more than one carrier relationship in place, with active-active or active-standby logic depending on traffic patterns.
Data plane isolation
Signalling and media planes are isolated from non-voice workloads so unrelated incidents cannot cascade into call quality issues.
Compliance & governance
FCC-approved, regionally aligned, operationally honest.
FCC approval
US operations are conducted under FCC approval, with processes aligned to current compliance requirements.
STIR / SHAKEN awareness
Caller identity attestation handled in line with US robocall mitigation expectations.
Regional alignment
Operational practices for UK, EU, Canada, Australia, and China reflect each market's expectations on number ownership and lawful use.
Lawful interception readiness
Where required by jurisdiction, lawful interception readiness is engineered into the platform — never bolted on.
Interoperability & quality
The work that keeps customers from noticing the network.
Continuous quality measurement
Every route is observed across ASR, MOS, jitter, and post-dial delay. Statistical shifts trigger investigation before customers feel them.
Interoperability testing
New customer endpoints and PBX systems are tested against documented SIP profiles before being placed in production.
Fault correlation
Incidents are correlated across signalling, media, and carrier metrics so root causes are diagnosed in hours, not days.
Change management
Routing changes are versioned, reviewed, and reversible. Production changes are never improvised at the keyboard.
Operational support model
How the network is actually run, day to day.
24 / 7 NOC
Continuous monitoring of the network with human response, not just alerting dashboards.
Tiered response
Severity-based response targets with engineering escalation when the issue warrants it.
Incident transparency
Customers affected by incidents receive direct communication, scope, and post-incident summaries.
Documented runbooks
Operational procedures are written, reviewed, and rehearsed — not improvised under stress.
Take the next step