Technical insight · Data Centre

In a critical facility the building is the smallest part of the cost.

A comms room or edge facility is a services problem with a building wrapped around it.

The constraint

Power, cooling, redundancy and the commissioning that proves them are the project. The structure is what holds them up.

A comms room or edge facility is a services problem with a building wrapped around it. The power path from the supply through switchgear, uninterruptible supply and distribution has to be designed for a redundancy topology, and the cooling has to be designed for a heat density that no ordinary commercial space produces.

Every one of those systems has to be proved to work, and proved to keep working when a component is taken out. That proof is a staged process that runs from the factory to a live load test, and it belongs in the programme as a work package with its own duration.

Redundancy

What you are buying when you buy resilience.

Redundancy is not a rating you select. It is a decision about how much of the system can be lost, or taken out for maintenance, while the load keeps running.

The three arrangements below cover most of what a small technical facility will consider. What separates them is not the number of components but whether the path is shared. A spare unit on a common board protects against a component failure and not against a board failure, which is the distinction that decides whether the room can be maintained without a shutdown.

Three redundancy topologies compared. Under N there is a single path from supply through switchboard and uninterruptible supply to the load, so any component failure takes the load down and maintenance requires a shutdown. Under N plus one there is a spare component on a shared path, so a single failure is survivable but the path itself is still common. Under 2N there are two entirely separate paths to a dual-corded load, so a whole path can be lost or maintained with the load still running. The cost rises with each step and so does what can be maintained without a shutdown.

N · one path

Supply

Switchboard

UPS

Load

Any failure stops the load. Maintenance means a shutdown.

N+1 · a spare on a shared path

Supply

Switchboard, still one

UPS

spare

Load

One component can fail. The path they share still cannot.

2N · two separate paths

Supply A

Supply B

Board A

Board B

UPS A

UPS B

Dual-corded load

A whole path can be lost, or maintained, with the load still running. It costs what it costs.

Three topologies and the honest consequence of each. The right answer is whichever one matches how long the business can be down, which is a question for the business rather than for the engineer.

Diagram of the staged commissioning process used on critical facilities: factory testing of equipment, site acceptance on delivery, component-level testing once installed, integrated systems testing across systems, and finally a load test that simulates real conditions and failure. Each stage has to pass before the next begins.

Staged commissioning: each stage passes before the next begins

  1. Factory testing

    before it ships

  2. Site acceptance

    on delivery

  3. Component test

    installed and alone

  4. Integrated test

    systems together

  5. Load and failure

    simulated for real

On a critical facility the building is not finished when it is built. It is finished when it has been proved, including proved to behave correctly when something fails. Commissioning is a work package, not a week at the end.

Staged commissioning on a critical facility. The final stage is the one that matters most and the one most often compressed, which is why we programme it explicitly.