Data centre power and UPS

Data centre power is a chain: utility supply, a standby generator, an automatic transfer switch, an uninterruptible power supply (UPS), and power distribution to each rack. The UPS bridges the gap between a utility failure and the generator taking over, and cleans up the power the rest of the time. Reliability comes from sizing each link correctly, removing single points of failure where the business needs it, and testing the whole chain regularly.

The power chain

1. Utility service arrives at the building's main electrical equipment. Larger sites sometimes have two feeds.
2. Standby generator starts automatically when utility power fails, usually within seconds.
3. Automatic transfer switch (ATS) moves the load from utility to generator and back. See automatic transfer switches.
4. UPS carries the IT load on batteries during the gap and conditions power continuously.
5. Power distribution units (PDUs) and rack PDUs deliver power to each piece of equipment.

Cooling must also stay running. A UPS that keeps servers powered while the cooling is off only buys a few minutes before equipment overheats. Decide what cooling equipment needs generator backup, and whether any needs UPS backup too.

Types of UPS

Double-conversion UPS systems give the cleanest, most consistent power and are the norm for data centres. Many have efficiency modes that trade some isolation for lower losses.

Batteries

Valve-regulated lead-acid (VRLA) batteries have been the standard for years. Lithium-ion batteries are increasingly common because they last longer, weigh less and take up less space, though they have their own fire safety and code requirements. Either way, batteries age, and heat shortens their life considerably. Battery rooms and cabinets need appropriate cooling, ventilation and fire protection as required by the applicable codes.

Runtime

If a generator is present, the UPS only needs enough runtime to cover generator start and transfer, plus a margin for a failed first start. Without a generator, the UPS needs enough runtime to shut equipment down cleanly, or to ride through the outages you expect.

Redundancy

Redundancy is described in a few common ways:

• N: just enough capacity for the load. Any failure or maintenance means downtime.
• N+1: one extra module or unit, so one can fail or be serviced without losing the load.
• 2N: two fully independent systems, each able to carry the whole load.

Most servers and network equipment have two power supplies. Feeding them from separate A and B paths, ideally from separate UPS systems, protects against a failure anywhere in one path. The Uptime Institute's tier classification is a widely used way to describe overall data centre resilience, from basic capacity to fully fault-tolerant.

A typical case: sizing a UPS for a small server room

Consider a server room with a few racks of servers, storage and network equipment, in a building with a standby generator. A sensible sizing process:

1. Measure the real load. Readings from existing rack PDUs or the current UPS are far better than nameplate ratings, which overstate actual draw. Record the load at a busy time.
2. Add planned growth. List the equipment expected over the life of the UPS, and estimate its draw from the manufacturer's typical figures.
3. Decide redundancy. For N+1, choose a modular UPS where the load fits within the capacity of all modules but one. For 2N, each of the two UPS systems must carry the full load alone.
4. Keep a margin. Size so the UPS is not running near its rating on day one. Many designers aim to leave meaningful headroom for growth and for the extra current when equipment starts.
5. Set the runtime. With a generator, runtime only needs to cover start, transfer and a retry if the first start fails. Without one, it needs to cover a clean shutdown of everything, including storage, which can take longer than expected.
6. Check the room. The UPS and batteries add heat and weight, need service clearance, and may need ventilation or fire protection depending on battery type and the applicable codes.

What a full power test should record

A simulated utility failure is the only way to prove the chain works end to end. Plan it with IT staff, the electrical contractor and the generator service provider, and record:

• Date, time, who was present and how the failure was simulated
• Load on each UPS and PDU before the test
• Generator start time and when it was ready to take load
• ATS transfer time and any alarms
• UPS behaviour: time on battery, battery voltage, any alarms or transfers to bypass
• Whether cooling restarted on its own and how high room temperatures rose
• Any IT equipment that rebooted or lost power, and why
• Retransfer to utility and the generator cool-down
• Actions arising, with owners and dates

Problems found in a planned test are far cheaper than problems found during a real outage.

Testing and maintenance

• Regular generator testing under load, including transfer through the ATS
• UPS preventive maintenance by qualified technicians, per the manufacturer's schedule
• Battery testing and replacement before end of life
• Infrared scans of electrical connections; see infrared thermography for electrical systems
• Load monitoring on each PDU and circuit, with alarms
• Periodic full tests that simulate a utility failure, planned with IT staff

Generator testing is covered in generator load bank testing.

Electrical work on UPS systems, switchgear and distribution must be done by licensed electricians and qualified service technicians, with lockout and arc flash precautions. Some generator, fuel storage and electrical installations also require permits and inspections under provincial rules.

How ehZee can help

Our data centre design and build service starts by asking what the room must support and how much downtime the business can tolerate, then plans power, cooling and layout together with licensed electrical trades. We also offer diesel generator inspections for existing standby systems.

Common questions