In the world of data centres, there is plenty of discussion about servers, networking, storage and cooling. Entire articles are written about network latencies, distributed storage architectures and energy efficiency. But there is a critical component that rarely makes headlines and yet is literally what separates a professional data centre from an improvised server room: the backup generator.
Generators are the silent heroes. They stand motionless in their rooms for months, sometimes years, waiting for the moment the electrical grid fails. When that moment arrives, they are the only barrier between service continuity and a catastrophic blackout that can cost clients thousands of euros per minute.
In this article we explain why generators are indispensable, what types exist, how they integrate into a data centre's power architecture, what the real response times are, how they are maintained and tested, what regulations govern them, and how fuel autonomy is sized to guarantee days of operation without mains power.
Why Generators Are Necessary
The public electrical grid is reliable, but not infallible. Even in developed countries with mature electrical infrastructure, power outages happen. Storms, substation failures, construction work cutting cables, demand spikes or incidents on the transmission network can leave an area without electricity for minutes, hours or even days. In Spain, the average duration of power supply interruptions is around 50-60 minutes per user per year, but a single prolonged outage can last hours.
A data centre hosting production servers, databases, SaaS platforms, hospital systems or financial services cannot afford even a single second of unplanned interruption. UPS (Uninterruptible Power Supply) systems cover the first seconds or minutes with batteries, but their autonomy is limited: typically between 5 and 30 minutes depending on the load and sizing. When UPS batteries run out, only the generator remains.
The generator is, therefore, the component that transforms a data centre from "we can survive a few minutes" to "we can survive for days". It is the difference between a Tier I and a Tier III or IV facility. It is what allows providers to sign availability SLAs of 99.99% with the confidence to deliver on that promise.
Types of Generators in Data Centres
Not all generators are the same. The choice of generator type depends on factors such as the power required, the location of the data centre, local environmental regulations and the available budget:
- bolt Diesel: the most widespread type in data centres worldwide. Diesel engines offer high reliability, fast start-up, high fuel energy density and a long service life. Available from 200 kW to several MW per unit. Their main disadvantage is particulate and NOx emissions, which are increasingly regulated.
- local_gas_station Natural gas: cleaner than diesel in terms of emissions, but requires a connection to the gas network, adding a second external point of failure. Less common in critical data centres as the sole backup source, although it is used in cogeneration configurations.
- swap_horiz Dual-fuel (diesel + gas): these combine the independence of diesel (fuel stored in local tanks) with the efficiency and lower emissions of natural gas. They start on diesel and can switch to gas once stabilised. They offer operational and economic flexibility.
- rotate_right Rotary UPS / Flywheel: hybrid systems that combine a flywheel with a diesel generator in a single unit. The flywheel provides instantaneous energy during the first seconds while the diesel engine starts and stabilises, eliminating the need for conventional chemical batteries.
Key fact:
Over 95% of data centres worldwide use diesel generators as their primary backup power source. Their proven reliability over decades, independence from locally stored fuel and autonomous start-up capability make them the industry standard.
Power Architecture in a Data Centre
Electricity in a data centre does not flow directly from the public grid to the server. It travels through a chain of components designed to filter, protect, switch and distribute power with maximum reliability. Understanding this chain is essential to grasping where and how the generator intervenes:
- electrical_services Utility feed: the connection to the public grid. Professional data centres typically have two independent feeds from different utilities or substations to reduce the risk of a simultaneous outage.
- switch_access ATS (Automatic Transfer Switch): the automatic switch that detects the loss of the primary feed and transfers the load to the generator (or to the second feed) within seconds. It is the brain of the switchover process.
- bolt Generator: starts automatically when the ATS detects a grid failure. In 10-30 seconds it reaches nominal speed and voltage and assumes the data centre load.
- battery_charging_full UPS (batteries): bridges the gap between the grid failure and generator stabilisation. It provides clean, uninterrupted power during those critical 15-60 seconds, plus an additional margin of minutes.
- device_hub PDU (Power Distribution Unit): distributes power from the UPS to each rack. Monitored PDUs allow real-time measurement of consumption per rack and per circuit.
- dns Rack and server: the final destination in the chain. Each server receives filtered, protected and backed-up power from the entire infrastructure above.
Power Redundancy Chain
The following table summarises the components of the power chain, their function, the time they cover and the typical redundancy level found in professional data centres:
| Component | Function | Bridge Time | Redundancy |
|---|---|---|---|
| Utility feed | Primary grid supply | Continuous (while grid is up) | N or 2N (dual feed) |
| ATS | Automatic switchover | 10-20 seconds | N+1 |
| Diesel generator | Backup power supply | Hours / days (tank dependent) | N+1 or 2N |
| UPS (batteries) | Seamless bridge | 5-30 minutes | N+1 or 2N |
| PDU | Distribution to rack | Continuous | A+B (dual feed) |
| Server PSU | Equipment power | Continuous | Dual PSU (A+B) |
Response Times: From Outage to Generator
When the electrical grid goes down, an automated sequence is triggered that requires no human intervention. Each step is timed to the second and the UPS covers the entire process so that servers experience no interruption:
- timer Outage detection (milliseconds): ATS sensors detect the drop in grid voltage or frequency within milliseconds. Simultaneously, the UPS switches to battery mode transparently.
- timer Start signal to generator (1-5 s): the ATS confirms the outage is not a transient micro-cut and sends the start signal to the generator.
- timer Engine start-up (5-15 s): the diesel engine starts, reaches nominal RPM and the alternator begins generating electricity.
- timer Stabilisation (5-10 s): the system verifies that the generator's voltage and frequency are within nominal parameters.
- timer Load transfer (10-20 s): the ATS transfers the data centre load from the grid (or UPS) to the generator. The transition is transparent to IT equipment.
Total time: 15-60 seconds. Throughout this entire process, the UPS feeds the load with battery power. Servers, switches, routers and cooling systems experience no interruption whatsoever. For the software running on the servers, the grid outage simply does not exist.
Maintenance and Testing
A generator that is not maintained or tested regularly is a generator that can fail when it is needed most. Preventive maintenance is as critical as the generator itself. Industry-standard protocols include:
- event_repeat Weekly no-load test: the generator starts and runs for 15-30 minutes without assuming a real load. It verifies that the engine starts correctly, that control systems function and that there are no alarms.
- event_repeat Monthly load bank test: a resistive load bank is connected to simulate the data centre's actual consumption. This prevents the problem of wet stacking (accumulation of unburnt fuel in the cylinders from always running at low load).
- event_repeat Annual full-load test: a complete simulation with a real transfer of the data centre load to the generator. It verifies the entire chain: ATS, generator, UPS, PDUs and the ability to operate for hours under real load.
- oil_barrel Fuel management: periodic diesel quality analysis (microbial contamination, water, degradation), fuel rotation and biocide treatment. Diesel can degrade in 6-12 months if not properly managed.
- thermostat Coolant and oil: verification of engine coolant levels and quality, oil and filter changes according to operating hours or calendar, inspection of hoses and belts.
Regulations and the Environment
Diesel generators do not operate in a regulatory vacuum. Data centres must comply with a series of regulations affecting emissions, fuel storage and noise:
- eco EU Stage V emissions: the European Stage V standard sets strict limits on particulate and NOx emissions for emergency diesel engines. Modern generators incorporate diesel particulate filters (DPF) and selective catalytic reduction (SCR) catalysts to meet these limits.
- volume_off Noise regulations: municipal ordinances set maximum noise levels in decibels according to zoning. Generators are installed with acoustic enclosures, hospital-grade exhaust silencers and vibration isolation to comply with limits even in urban environments.
- propane_tank Fuel storage: diesel tanks must comply with chemical products storage regulations, with bunds of at least 110% of the tank volume, leak detection, earthing and fire protection.
- description Licences and permits: classified activity permit, environmental authorisation, emergency and contingency plan, periodic inspection by authorised control bodies and electrical installation review.
Fuel and Autonomy
A generator's autonomy depends on two factors: the engine's consumption at the actual data centre load and the fuel tank capacity. A 1 MW diesel generator consumes approximately 250-280 litres of diesel per hour at 75% load. These numbers allow tanks to be sized to cover the most critical scenarios:
- local_gas_station 24-hour autonomy: requires approximately 6,500-7,000 litres of diesel for a 1 MW generator. This is the minimum recommended for production data centres.
- local_gas_station 48-hour autonomy: between 13,000 and 14,000 litres. Standard for Tier III data centres that need to cover an entire weekend without refuelling.
- local_gas_station 72-hour autonomy: close to 20,000 litres. Tier IV level, designed to cover prolonged blackouts with margin to coordinate additional fuel deliveries.
During prolonged outages, refuelling logistics become a critical factor. Professional data centres maintain contracts with fuel distributors that guarantee priority delivery in emergency situations, even when the electrical grid fails across an entire region. Planning access routes for tanker trucks, quick-connect points to the tank and hot-refuelling protocols (without stopping the generator) are all part of the contingency plan at any professional facility.
N+1 and 2N Redundancy
Having one generator is necessary. Having more than one is what defines a truly reliable data centre. Industry standards (Uptime Institute Tier Classification) establish two main levels of redundancy in the power infrastructure:
- add_circle N+1 (Tier III): if the data centre load requires N generators to operate, one additional spare is installed. If a generator fails or is undergoing maintenance, the spare generator assumes its load automatically. It allows concurrent maintenance: a generator can be taken offline for servicing without affecting the service.
- looks_two 2N (Tier IV): the entire power chain is fully duplicated. Two independent utility feeds, two sets of generators, two UPS systems, two PDUs per rack, two power supplies per server. If one complete chain fails (for any reason), the other assumes 100% of the load. There is no single point of failure in the entire power infrastructure.
Practical difference:
In an N+1 configuration, one generator can be lost without consequences. In a 2N configuration, an entire power chain (utility feed + generator + UPS + PDU) can be lost without a single server losing power. The choice between N+1 and 2N depends on the criticality of the service and the committed SLA.
EasyDataHost: Generators at Our Data Centre
EasyDataHost's infrastructure is housed in a Tier III+ data centre in Madrid with a power chain designed to eliminate every single point of failure. The facility features N+1 redundant diesel generators with automatic start-up, double-conversion UPS with extended battery autonomy, dual independent utility feeds and redundant A+B PDUs in every rack.
Generators are tested weekly without load and monthly with a load bank. Fuel tanks are sized for a minimum autonomy of 48 hours, with priority supply contracts for emergency refuelling. The entire power chain is monitored 24/7 with automatic alerts for any anomaly, as part of our managed services.
- check_circle N+1 diesel generators: redundancy with automatic start-up in under 15 seconds.
- check_circle Double-conversion UPS: uninterrupted power during the grid-to-generator switchover.
- check_circle Dual utility feeds: two independent feeds from the public grid.
- check_circle 48h+ autonomy: tanks sized to operate for days without mains power.
If you are looking for a data centre in Spain with professional-grade power infrastructure for your colocation or dedicated servers, you can view our facility details or contact our team.
Conclusion
Generators are not glamorous. They do not appear on cloud service spec sheets or in marketing brochures. But they are, quite literally, the last line of defence between service continuity and a total blackout. Without reliable, well-maintained and properly redundant generators, no data centre can guarantee the availability levels that enterprise production demands.
- arrow_right Diesel generators are the backup standard in over 95% of data centres worldwide.
- arrow_right The UPS covers the critical 15-60 seconds between the outage and generator stabilisation.
- arrow_right Preventive maintenance (weekly, monthly and annual tests) is as critical as the generator itself.
- arrow_right N+1 or 2N redundancy eliminates the single point of failure in the power chain.
- arrow_right EasyDataHost operates with N+1 generators, dual utility feeds and 48h+ autonomy in Madrid.
Next time you see a 99.99% SLA, remember that behind that figure there is a diesel generator waiting in silence, ready to start within seconds and keep your services online for days. If you want to learn about our power infrastructure in detail, get in touch.