Bank Achieves 99.98% Uptime on OCI

A bank running customer facing services on Oracle Cloud Infrastructure had a requirement that admitted no compromise: the services had to be available, because in banking an outage is not an inconvenience but a breach of trust and often a regulatory matter. This anonymized case study describes how the bank's estate was designed to deliver 99.98 percent uptime, a level its own data centres had never reached, and how that availability was sustained rather than merely achieved once. The client is anonymized by sector, but the design and the numbers are real.

It is one of the case studies behind our OCI case studies and benchmarks pillar, and the availability it describes sits within the range documented in OCI uptime benchmarks by workload.

The situation

The bank's services had previously run in its own data centres, where availability had been a constant struggle. Building genuine resilience on premises meant duplicate hardware, a second data centre, and tested failover, each of which was expensive and complex, and the result was an estate that was resilient in theory but rarely tested in practice. Outages, when they came, were painful and slow to recover from, because the failover that was supposed to save the day had not been exercised often enough to be trusted.

Moving to OCI was an opportunity to design resilience properly rather than carry forward the compromises the data centre had forced. The bank set a clear target, availability that would comfortably exceed what it had managed before, and the estate was designed from the start to meet it. Crucially, the goal was not a one time achievement but a sustained level, which shaped every decision toward designs that could be tested and maintained.

The design: no single point of failure

The core principle was that no single failure should be able to take a service down, which on OCI is a design choice rather than a capital project. The application tiers were spread across fault domains within an availability domain, and for the most critical services across availability domains, so that the loss of any single piece of underlying infrastructure left the service running on the rest. Load balancers health checked the application instances and routed only to healthy ones, so a failed instance was removed from service automatically.

The databases, the part of the estate where availability is hardest and most important, were protected with standbys kept continuously synchronised, so that the failure of a primary database could be answered by promoting its standby rather than by a slow restore. This layered design, set out in our disaster recovery and HA solution, meant that every component had a peer ready to carry its load, which is the foundation of high availability.

Tested failover, not theoretical failover

The lesson the bank had learned painfully on premises was that untested failover is not failover at all, because the failover that has never been exercised is the failover that fails when it is finally needed. The OCI design was therefore built to be tested, and the failover was rehearsed on a regular schedule rather than assumed to work. Each rehearsal confirmed that a primary could fail and its standby take over within the time the bank required, and each one surfaced and fixed the small issues that always accumulate.

This discipline, treating failover as something to be proven repeatedly rather than trusted blindly, is what separated the bank's OCI estate from its old data centres. The technology to fail over existed in both, but only on OCI was it exercised often enough to be relied upon. Regular tested failover is the single practice that most distinguishes estates that achieve their uptime targets from those that merely hope to.

Continuous monitoring

High availability is not only about surviving failures but about seeing them coming, which is why continuous monitoring was built into the estate from the start. The monitoring watched not just whether services were up but the signals that precede trouble, rising latency, growing error rates, capacity approaching its limits, so that problems could be addressed before they became outages. Alerts were tuned to fire on the signals that genuinely mattered rather than on noise, so that an alert meant action.

This combination of resilient design and vigilant monitoring is what turned a high availability target into a sustained reality. The design ensured that failures did not cause outages; the monitoring ensured that the conditions leading to failures were caught early. Sustaining that over time is the work of our managed services, because availability is a continuous achievement rather than a one time build.

The measurable result

The estate delivered 99.98 percent uptime, comfortably exceeding the bank's target and far surpassing what its own data centres had managed. More importantly, the availability was sustained across the period the estate ran, because it came from a design that survived failures and an operation that caught problems early, rather than from luck. The bank gained not just a number but confidence, the freedom to operate knowing that a single failure would not become a customer facing outage.

The 99.98 percent figure sits within the range achievable for well designed estates documented in our uptime benchmarks, and the cross region dimension of resilience is explored in the healthcare cross region DR case. The lesson is the one that runs through our case studies pillar: availability of this order is designed and maintained, not bought, and any estate willing to design for it can reach it.