A reliable system delivers correct outputs under specified conditions. A resilient system delivers acceptable outputs under unspecified conditions. In an environment where the conditions are predictable—where traffic patterns, failure modes, and load profiles are known—reliability is sufficient. In an environment where conditions change unpredictably—where a geopolitical event reroutes traffic, a supply chain disruption changes data patterns, or a regulatory change alters processing requirements—resilience is essential.
Reliability vs. resilience. Reliability asks: "Will the system work as designed?" Resilience asks: "Will the system produce acceptable outcomes even when the design assumptions are violated?" These are fundamentally different questions. A system that is 99.99% reliable under normal conditions but 0% functional under abnormal conditions is less resilient than a system that is 99% reliable under normal conditions and 95% functional under abnormal conditions.
Designing for resilience. Resilient systems share several properties: graceful degradation (the system continues to provide reduced service rather than failing completely), circuit breaking (the system stops trying operations that are repeatedly failing), bulkheading (the system isolates failures so they do not cascade), and adaptive capacity (the system can reconfigure itself in response to changed conditions). These are not features that can be added after the fact—they are architectural properties that must be designed in.
The government context. Public-sector systems face a unique set of unpredictable conditions: sudden regulatory changes (a new EU directive requires immediate data processing changes), geopolitical events (sanctions require immediate service restrictions to specific countries), and demand spikes (crisis events increase service load by orders of magnitude). These conditions cannot be predicted, but they can be prepared for. EU government tenders on TED increasingly specify resilience requirements—not just uptime SLAs, but recovery time objectives, degraded service modes, and incident response procedures.
Building adaptive capacity. The most important property of a resilient system is adaptive capacity—the ability to reconfigure under pressure. This requires: configuration-driven behavior (so changes can be made without code deployment), operational playbooks for known failure modes (so operators do not have to improvise under stress), and monitoring that detects anomalies before they become outages (so the system can adapt proactively rather than reactively).
Reliability is a baseline. Resilience is the goal. The systems that perform best in the real world are not those that never fail—they are those that adapt when failure is inevitable and continue delivering value even when conditions are far from specified.