Executive Summary
- A data centre needs time like it needs power, says Elena Parsons, Strategic Business Manager, Timing Services, National Physical Laboratory (NPL).
- And, timing should be planned in from the start, as a layered system.
- But, the starting point is understanding your timing dependency, where does it come from, who owns it, and how is it traced?
For every critical dependency in a data centre, there is a backup, and often a backup for the backup. Power has Uninterruptable Power Supply (UPS) and generators. Cooling has redundant chillers. Connectivity has diverse fibre routes. Physical and cyber security are layered by design. That is how operators deliver the five nines – 99.999% uptime – that customers now expect as standard.
But one dependency rarely gets the same treatment: time.
Why every data hall depends on time
A data centre needs time like it needs power. Every server, switch and storage array keeps its own clock, and left to itself, each one drifts. So all of them are constantly steered to a common time reference.
Trusted timestamping is crucial; this shared sense of ‘now’ is what enables databases to process transactions in the right order, assists recovery from corrupted backups, and allows thousands of machines to work as one system.
The demands on that shared reference are growing. AI training relies on tightly synchronised Graphic Processing Units (GPUs), and AI applications draw on timestamped data from many locations at once.
As operators spread across campuses and edge sites in search of power and cooling, every new site adds clocks that must stay aligned with the rest.
Customer Service Level Agreements (SLAs) increasingly call for trusted timestamping for transaction logging, audit and compliance. Timing disruptions that cause equipment to reboot can create hotspots, wasting energy that is already in short supply.
A single point of failure on the roof
In most facilities, time is derived from satellite signals, like Global Navigation Satellite Systems (GNSS), via an antenna on the roof. Satellite timing is accurate, ubiquitous and cost-effective, meaning it has become a ready foundation for timing services in many data centres.
Jamming, space weather, local interference or equipment failure, however, can all disrupt the signal. Holdover oscillators will help to keep clocks close for a while, but the length of time they can do so accurately depends on the quality of the oscillator, with differences building up within hours or days. Spoofing, where fake signals trick a receiver into reporting the wrong time, can do damage immediately, and may not be noticed.
Few co-location providers offer a timing service, so tenants typically deploy their own GNSS-based timing system. Roof space is limited and antenna feeds are often shared, so one faulty antenna can block time for many customers at once – a fault no individual tenant can see, let alone fix.
In enterprise data centres, run by small, generalist teams, timing often has no owner at all.
Specify timing at the design stage
Part of the problem is where timing sits in the design process. Power and cooling are specified in detail on every Mechanical, Electrical, and Plumbing (MEP) scope.
Timing rarely appears and is often treated as the customer’s responsibility or bundled in with generic Network Time Protocol (NTP) services. So it gets bolted on later, when changes are harder and retrofits cost more.
Timing should be planned in from the start, as a layered system: GNSS supported by complementary terrestrial timing signals accessed through diverse dissemination routes, so each can back up and cross-check the others. This makes the chance of losing synchronisation far less likely.
The UK Government has invested in the National Timing Centre (NTC) programme to develop a sovereign, resilient timing network. This will provide data centre operators with timing signals from a range of distribution technologies including fibre, communication satellites, and terrestrial broadcasts.
These can be used alongside existing Global Navigation Satellite Systems (GNSS), giving greater resilience and confidence in service continuity and performance.
Layered timing resilience
For all types of data centres, including co-location tenants, traceable timing provides assurance and supports compliance, strengthens cyber security, improves auditability and reinforces trust in the digital services that organisations and their customers rely on every day.
For co-location providers, there is also an opportunity to enable layered timing resilience for tenants by becoming a delivery partner for terrestrial timing, distributing assured time as a value-added service alongside core services like power and connectivity. Many customers, across telecoms, financial services and energy, need traceable time for compliance, so this could become a new revenue stream, while sparing tenants the cost of deploying dedicated timing links of their own.
Question your timing provision
The starting point is understanding your timing dependency. Where does each system get its time from, and how is that timing delivered? What happens if that source fails, or is wrong? Who owns timing in the facility? And can the provider show that its time is traceable to UTC?
Enterprise IT teams using cloud or co-location services should be asking their providers the same questions, and the Royal Institute of Navigation’s Resilient PNT Checklist is a useful guide.