Designing a data centre for a public institution

What does a public institution's data centre need to do?

It needs to keep the institution's critical services running through power cuts, equipment failures and maintenance. For a hospital that may be patient records and imaging. For a university, the campus network, student systems and research storage. For a government body, citizen services and surveillance video.

The design starts with a list of those services and how long each can be down. That list sets the redundancy, which sets the cost. Standards help structure the choice. TIA-942 covers data centre infrastructure, including cabling, power, cooling and levels of redundancy, and the Uptime Institute's Tier classification is widely used to describe availability. A buyer does not need the highest level everywhere; it needs the right level for each service.

How should power be designed?

Power should be designed so that no single failure stops the critical load. That means a path from the utility to each server with backup at every stage.

• Utility feed, ideally from two sources where available
• Diesel generator (DG) sized for the full load, with an automatic transfer switch and fuel for an agreed run time
• UPS to carry the load through the gap before the generator starts, often in N+1 or 2N configuration
• Power distribution units (PDUs) in each rack, with dual feeds for equipment that has dual power supplies
• Earthing and surge protection designed with the building, not added later
• Metering at each level, so capacity and efficiency can be tracked

What does precision cooling involve?

Precision cooling means air conditioning built for IT equipment: it runs continuously, controls humidity as well as temperature, and is sized for the heat the racks produce. Comfort air conditioning built for offices is not designed for this and should not be used for the white space.

Airflow matters as much as cooling capacity. Racks are arranged so that cold air enters at the front and hot air leaves at the back, in alternating cold and hot aisles. Containment closes off one aisle so the two air streams do not mix. For dense racks, in-row cooling places cooling units next to the load. Cooling units should have redundancy just like power, since overheating can stop equipment as fast as a power cut.

Why does structured cabling matter so much?

Structured cabling decides how easy the data centre is to run, change and fault-find for its whole life. It is also the hardest part to change once racks are live.

A structured design uses a planned hierarchy: backbone cabling, usually fibre, between main and zone distribution areas, and horizontal cabling to each rack. Cable runs in overhead or underfloor trays kept separate from power. Every port and cable is labelled and recorded. Spare capacity is built in so new equipment does not mean new cable pulls through a live room.

What is DCIM, and is it needed?

DCIM is software that monitors and records the physical side of a data centre: power draw, UPS and generator status, temperature and humidity, leak detection, door access and rack space. For a public institution with a small IT team, it is often the difference between catching a failing UPS battery on a dashboard and finding out during a power cut.

At minimum, DCIM should alert on power, cooling and environmental alarms, track capacity, and keep a history for audits. It works best when the power and cooling equipment is chosen with open monitoring interfaces from the start.

How should disaster recovery be planned?

Disaster recovery is planned from two numbers the institution agrees on for each service. The recovery point objective (RPO) is how much data it can afford to lose. The recovery time objective (RTO) is how quickly the service must be back. A short RPO means frequent or continuous replication; a short RTO means standby systems ready at the second site.

The DR site should be far enough away not to share the same risks, such as the same flood zone or power grid. Backups should include at least one copy that cannot be altered or deleted by an attacker. Most importantly, recovery should be tested on a schedule. A DR plan that has never been run is an assumption.

How Ahuva approaches data centre work

Ahuva expanded into data centres in 2024, offering end-to-end design, implementation and management. Its scope covers design, server, storage, virtualisation and HCI; power, precision cooling and structured cabling; DCIM and high-availability architectures; and backup, disaster recovery and business continuity. Its OEM authorisations include Vertiv for data-centre power and precision cooling, HPE for server, storage and compute, and Cisco for network and compute.