Technical comparison of backup solutions for hospitals, data centers and critical sites, based on the expected level of availability.
Introduction
When continuity of operations is non-negotiable, the emergency power strategy is one of the first decisions to get right. Two technologies dominate the conversation: backup generators and uninterruptible power supplies (UPS, called ASI in French). They are often presented as alternatives, but on most critical sites they are complementary.
This article explains what each technology actually does, where each one is appropriate, and why hospitals, data centers and industrial facilities typically deploy both together in a coordinated architecture.
Why this topic matters
The cost of a power interruption is rarely limited to the outage itself. It cascades into unsafe conditions for people, damaged equipment, lost production, corrupted data and, in regulated environments, contractual and regulatory exposure.
Choosing the right emergency architecture is not a purchasing exercise. It is an engineering decision that must reflect the site's actual loads, tolerance to voltage dips, minimum autonomy requirements and the operational reality of maintenance and testing.
Technical context
A UPS interposes stored energy — typically batteries — between the mains and the load. In an online (double-conversion) topology, the load is permanently supplied through the inverter, so a mains loss is invisible: the transfer time is effectively zero and the output waveform stays clean.
A backup generator is a rotating machine — most often a diesel engine driving an alternator — that starts on loss of mains, stabilizes, and then takes over the load through a transfer switch. Between the mains loss and the moment the genset is stable and connected, several seconds pass. That gap is unacceptable for sensitive electronics.
The two technologies therefore operate on very different timescales: milliseconds for a UPS, several seconds to tens of seconds for a genset, but minutes for a UPS's battery autonomy versus hours or days for a genset with a fuel reserve.
Key engineering considerations
Choosing between — or combining — these technologies comes down to a small number of engineering questions:
- Transfer time tolerance: can the load survive a short break, or does it require a bumpless supply?
- Required autonomy: minutes to allow a controlled shutdown, hours to ride through a local incident, or days to remain operational during an extended grid event.
- Load profile: constant, cyclic, motor-heavy, or dominated by power electronics.
- Segregation of critical circuits: which sub-panels genuinely need to be on the emergency bus and which can be shed.
- Redundancy strategy: N, N+1 or 2N architectures, driven by acceptable failure modes rather than by catalogue.
- Maintenance and testing constraints: access to the plant, fuel logistics, battery lifecycle and the possibility of running load-bank tests.
Practical recommendations
For IT rooms, medical imaging, laboratory instruments and process control, a UPS is the primary line of defence. It absorbs voltage dips, filters disturbances and gives the site time to react in a controlled way.
For life-safety systems, HVAC serving critical rooms, refrigeration, and any load that must remain available for more than a few minutes, a generator is the appropriate answer. Its role is to sustain the site once the initial event has passed.
On sites where both categories of load exist — which is the case for most hospitals, data centers and modern industrial facilities — the two technologies are wired in series. The UPS carries the sensitive loads from the instant the mains fails; the generator starts, stabilizes and then supplies the UPS's rectifier and the non-UPS emergency circuits. Batteries are sized to cover the genset start-up window with margin.
Redundancy is a design choice, not a marketing claim. N+1 on gensets and UPS modules is a common baseline for high-availability sites. Higher levels (2N, 2N+1) are reserved for architectures where any single fault must remain invisible to the load.
Common mistakes to avoid
A UPS is not a long-term power source. Sizing it for hours of autonomy is almost always the wrong answer; it inflates cost, footprint and battery maintenance without solving the real problem.
A generator alone does not protect sensitive electronics from the transfer gap or from routine grid disturbances. Deploying only a genset on a data-driven site guarantees incidents on every mains event.
Skipping periodic load-bank testing is one of the most common — and most damaging — omissions. An untested emergency chain is an assumption, not a capability.
Under-sizing the emergency bus by ignoring future loads leaves the site unable to accommodate the growth it is designed to enable. The emergency architecture must be planned with the same horizon as the rest of the installation.
How EOS approaches this topic
We start from the site's actual load inventory and its business continuity requirements, not from a preferred product. That means mapping every circuit, classifying it by criticality, defining acceptable transfer times and target autonomies, and only then selecting the technology mix.
Our engineers design the coordinated architecture — UPS topology, genset sizing, transfer scheme, fuel and battery strategy, monitoring — and specify the commissioning and periodic testing procedures needed to keep it credible over time.
Related solutions
The following EOS solutions address the topics discussed above and are commonly combined on the same project.
Discuss your requirements
If you are planning a new facility, renovating an existing one, or reviewing an emergency architecture that has grown over time, our team can help you clarify the technical options and their operational implications.