HA-NVS

The Problem

NIMC needed to verify National Identification Numbers in real time against a national identity database that keeps growing past 250 million records, with government agencies, banks, telcos, and other licensed partners all depending on it staying fast and available.

Constraints

• ▹Verification requests had to resolve in under a second, even as the source database grows past 250M+ records.
• ▹99.99% uptime, 24/7, sitting in the critical path for other systems, including billing.
• ▹Verification nodes are physically distributed across authorized partner data centers, not one central location.
• ▹Has to comply with the OSIA identity standard and NIMC data protection policy.
• ▹As a government identity system, policy required data like enrolment files to stay hosted entirely in-house rather than on public cloud storage, ruling out the usual managed-storage shortcuts.

How It Was Solved

• ▹Split into six purpose-built services: an ETL pipeline, a verification record store, a history/audit store, a messaging engine, a gateway managing the distributed nodes, and the verification nodes themselves, so no single service had to do everything.
• ▹An event-driven messaging layer (built on Kafka) propagates enrolment changes out to distributed nodes so each one can serve verifications locally instead of round-tripping to a central server.
• ▹New enrolments and edits replicate to every node within 24 hours, trading a small propagation delay for consistent sub-second reads under load.
• ▹File data (enrolment documents, biometric artifacts) had to stay in-house per policy, so instead of reaching for a managed object store, we built a dedicated file service on Spring Boot backed by MongoDB, running entirely on our own infrastructure.
• ▹Enrolled citizen data is sourced from NIMC's national identity database over Oracle, the existing system of record, rather than duplicating it elsewhere.