A citizen-facing service that fell over at every deadline peak, and partner integrations handed out as raw database access. We rebuilt the edge and put every partner behind a gateway.
Client: Public sector agency (anonymized)
−40%
Downtime post-launch
99.95%
SLA sustained over 6 months
0
Direct database partners left
Challenge
Peak-day traffic caused cascading failures; no autoscaling and a single AZ.
Partner ministries integrated through direct database reads with shared credentials.
No SLA measurement, so outage duration was disputed after every incident.
Approach
Kong gateway with mTLS, per-partner quotas, signed payloads and full request audit.
Multi-AZ AWS deployment, read replicas, queue-backed writes for peak absorption.
Chaos and load testing against real deadline-day traffic shapes before cutover.
SLA instrumentation with an agreed measurement definition published to partners.
DR runbook with quarterly tested failover and a named owner per service.
Architecture
flowchart LR
C[Citizens] --> E[CDN + WAF]
PM[Partner ministries] -->|mTLS + quota| K[Kong API Gateway]
E --> N[Node service tier - multi AZ]
K --> N
N --> Q[[Write queue]]
Q --> DB[(Postgres primary)]
DB --> RR[(Read replicas)]
N --> O[Observability + SLA metering]
Every partner enters through the gateway; writes queue so a traffic spike never reaches the database directly.
Stack
AWS
Kong
Postgres
Node
Bring the initiative that is stuck. We will tell you what we would do in the first two weeks.
Thirty minutes, an architect on the call, no deck. You leave with a position on sequencing and risk whether or not you hire us.