Critical national infrastructure has no room for delayed decisions. When a smart metering programme connects 53 million devices across 30 million homes and businesses, every message, credit update, and fraud check has to be correct the first time. Legacy database architectures were never built for this kind of pressure, and when authentication and anti-replay checks lag behind, the risk isn’t just downtime. It’s duplicate credits, unauthorised access, and safety consequences for millions of households.
This case study breaks down how CGI, the Data Services Provider for the UK’s national smart metering programme, replaced a disk-bound Oracle architecture with a real-time decisioning layer capable of authenticating tens of millions of messages a night at sub-10 millisecond latency. You’ll see how a national infrastructure operator moved from batch reconciliation to authoritative, in-transaction decisions, and what that shift meant for availability, fraud prevention, and operational resilience.
At a high level, the approach centers on keeping all active request state inside a single real-time decisioning engine rather than spreading authentication, queuing, and validation logic across separate systems. Southbound commands are authenticated and authorised before they ever reach a device. Northbound responses are correlated against their originating requests in the same tier. Active-Active replication across two data centres keeps both sites processing simultaneously, so a regional failure becomes a capacity event rather than an outage.
Whether you’re responsible for critical infrastructure reliability, fraud prevention, or platform architecture at national scale, this case study shows what happens when decision authority moves into the data layer instead of trailing behind it. CGI went from three-nines availability and 4 to 6 hour maintenance windows to four-nines uptime and zero-downtime upgrades, without adding headcount or infrastructure. Read on to see how the architecture works in practice.