TECH & AI
Uptime has become the defining performance metric of the modern data centre. Digital services now underpin everything from financial markets to transport systems, and tolerance for disruption has all but disappeared. In mission-critical environments, 99.999 % availability and five minutes of downtime a year form the baseline.
Industry performance is improving. The Uptime Institute’ s Annual Outage Analysis 2025 records a fourth consecutive year of improved service availability. High-profile outages continue to show how fragile uptime becomes when resilience is applied unevenly across infrastructure layers, and that unevenness often traces back to how those layers are designed and prioritised.
“ At five-nines scale, an automationfirst design is the foundation for delivering the level of resilience modern digital infrastructure demands”
Ciaran Flanagan VP and Global Head of Data Centre Solutions
& Services Siemens
Earlier this year a power substation failure disrupted operations at Heathrow Airport, stranding passengers and cargo. A separate US cloud outage hit communications, commerce and entertainment platforms used by millions. Both incidents point to the same weakness: even well-designed infrastructure can still go offline if a single point of failure remains.
Ciaran Flanagan, Vice President and Global Head of Data Centre Solutions & Services at Siemens, argues that the control layer governing those systems is routinely left out of redundancy planning.
In this Q & A with Data Centre Magazine, he sets out where duplication stops working, why controls belong at the start of the design process and what automation-first design means for facilities operating under rising availability expectations.
Q. WHERE DOES REDUNDANCY ALONE REACH ITS LIMITS?
» For decades, the industry’ s primary response to uptime risk was duplication. The 2N model – two of every critical system – became the benchmark for high-availability facilities. Power, cooling, fire protection, and security systems were mirrored so that failure in one path could be absorbed by another.
This approach raised the baseline for reliability but wasn’ t infallible. Incidents such as cooling failures at large colocation facilities have shown how faults can cascade across both primary and backup systems, halting operations even in environments designed for resilience.
54 September 2026