← All articles
5 min read

Inside a modern datacentre

Every file you upload travels through buildings most people will never see. Modern datacentres are remarkable feats of engineering - designed to keep servers running continuously, efficiently, and reliably despite power cuts, equipment failures, and the relentless heat generated by thousands of machines.

99.999% Tier IV uptime SLA
1.2–1.5 Target PUE ratio
2N Power redundancy

Power: the first concern

A medium-sized datacentre can consume as much electricity as a small town. Power is the primary operational cost and the primary failure risk. Enterprise facilities address this with layered redundancy:

Tier ratings (defined by the Uptime Institute) classify datacentres from Tier I (basic, ~99.67% uptime) to Tier IV (fully fault-tolerant, ~99.995% uptime). The difference between Tier III and Tier IV is whether maintenance can be performed without any downtime at all.

Cooling: the constant challenge

Servers convert electricity into computation - and heat. A single rack of modern servers can generate 10–30 kW of heat, equivalent to dozens of electric heaters. Removing that heat efficiently is one of the defining engineering challenges of datacentre design.

Traditional air cooling uses computer room air conditioning (CRAC) units to push chilled air beneath raised floors. Hot air rises from servers into a "hot aisle" and returns to cooling units in a closed loop. Modern high-density deployments increasingly use direct liquid cooling - cold plates on CPUs, or even full immersion in dielectric fluid - to handle heat densities that air simply can't manage.

Cooling efficiency is measured as PUE (Power Usage Effectiveness): total facility power divided by IT equipment power. A PUE of 1.0 would be perfect - all power goes to servers. Real-world figures range from 1.2 (excellent) to 2.0+ (legacy facilities with poor cooling design).

Connectivity: the nervous system

A datacentre without network connectivity is just an expensive heated building. Carrier-neutral facilities attract multiple internet service providers, creating a competitive environment for bandwidth pricing and ensuring that no single provider failure isolates the facility. Physical fibre enters through diverse conduit routes - often via different streets - to eliminate single points of failure.

Within the facility, high-speed switching fabric connects servers at 25, 100, or 400 Gbps. Cross-connects allow direct physical connections between tenants or to carriers, bypassing the public internet for ultra-low latency exchange.

Physical security

The physical security of a datacentre is often more rigorous than most office buildings. Biometric access controls, man-traps (double-door vestibules that prevent tailgating), 24/7 security personnel, extensive CCTV coverage, and strict visitor logging are standard at enterprise facilities. Cage systems within the floor allow different tenants to maintain private, locked spaces even within a shared facility.

Geographic selection matters. Good datacentres are sited away from flood plains, flight paths, and seismic fault lines. Proximity to renewable energy sources - hydroelectric in the Pacific Northwest, geothermal in Iceland - increasingly factors into both cost and sustainability decisions.

← All articles