Case Studies Public sector
Leading MEA border control
System Twin — Border Control (MEA)
System twin for border operations—e-gate and counter staffing against passenger load; Kafka Streams event-sourced rebuild; dual-write consistency; 2.5DC stretch cluster; e2e p99 under 50ms at zero RTO/RPO.
Apache Kafka · Kafka Streams · Kubernetes
-
Customer
A leading Middle East & Africa border-control operator. Mission-critical passenger flow: e-gates, counters, staff and shifts under peak arrival waves.
The system had to stay up for operational simulation and surveillance—matching capacity to load—at zero RTO / RPO.
-
Challenge
Simulate and monitor e-gate, staff, and counter availability against arriving passenger load.
The incumbent stack was largely Spring Boot with request-reply bolted onto Kafka. That shape had serious performance problems. The estate was also a 2.5DC stretch cluster with known gaps—notably Zookeeper timeouts—that blocked a trustworthy latency envelope on a zero RTO/RPO system.
-
Work
Owned system architecture and performance engineering. Rebuilt the core as a domain-modeled twin: event-sourced aggregates for system state and simulation, on Kafka and Kafka Streams.
Moved off request-reply-on-Kafka. Dual-writes to Kafka alongside the RDBMS kept consistency without CDC overhead. Hardened the 2.5DC stretch cluster—including Zookeeper timeout failure modes—so the latency envelope could hold.
-
Impact
- End-to-end p99 under 50ms on the decision / simulation path
- Event-sourced twin for e-gate and counter capacity against passenger load
- Dual-write consistency without CDC overhead; stretch-cluster gaps closed for zero RTO/RPO
