What’s the Problem?
Temperature data from the UK Met Office is prompting thoughts on areas where data centre equipment, facility design or operation may have to change to cope with changing weather conditions.
The Met Office overview of the UK having a temperate maritime climate in general means that we have a cool and mild climate. But there appear to be more instances of heat records being broken.
Conventionally, August 2003 was taken as having the hottest day in recent UK history, with over 38C in Faversham, Kent. But in 2022, ambient temperature went over 40C in Coningsby, Lincolnshire (see Figure 1). The Met Office recorded this February as the warmest on record for England and Wales. So, monthly annual averages can be useful in showing the min/max range, but of course bely the extremes.
Figure 1 UK hottest day data Figure 2 Example of ambient temperature data
Data centre organisations can provide their operations teams with warnings derived from holistic views of daily, monthly and annual average temperature data at a national or regional level. Such data may also assist designers and planners, and be useful in discussions with funders and local authorities. The UK Met Office or the European Copernicus programme are helpful sources.
In terms of outlook, the Met Office had 2024 as potentially a further record-breaking year globally. Indeed, 2024 could be the first time we see values of 1.5 °C or above versus the benchmark 1850-1900. And, across the UK as a whole, the Met Office expects to see:
Warmer winters – by 2070 between 1 and 4.5°C warmer than 1990
Hotter summers - by 2070 between 1 and 6°C warmer than 1990, depending on the region
More frequent and intense weather extremes
So temperature shifts pose significant challenges for data centre equipment, and operations teams. For example, as ambient temperatures rise, cooling systems may struggle to maintain optimal server temperatures within existing data centres. In turn, that could lead to server overheating and outages, potentially disrupting services, and risk data loss if servers and hard drives fail. The underlying risk of fire, with the harm that could bring, is always a priority. And the impacts of high/extreme temperatures to people’s health and productivity is a priority (see for example the UK Government’s periodic Climate Change Risk Assessment report).
When higher cooling loads are translated into total power demands, summer power disruptions might become as important as winter power disruptions. With data centres often being part of a local metro ecosystem, the regional power supply could struggle to cater for data centres along with businesses and homes. Failure of the electricity transmission, nationally and regionally, is considered in the UK Government’s National Risk Register which can be useful for data centre operators to engage with.
Tackling rising and even extreme ambient temperatures to keep existing data centres operating reliably and efficiently day by day, year on year, is quite an evolving challenge. New data centres or expansions of existing facilities falls into the same challenge.
What can be done?
At corporate level, weather or more specifically temperature, could be elevated to be a priority risk. That allows a company-wide assessment of the robustness of monitoring per equipment item checking thermal performance. This may identify a need for further temperature sensors and enhancing DCIM configuration, combined with improved dashboarding and management reporting. Linking in incident reporting and root cause analysis starts to give a holistic view of thermal performance versus the risk. As well as supporting operations teams, such data can also be valuable when fed back into modelling for data centre expansions or planning of new facilities.
Extreme heat can increase equipment failure rates within data centres. New investment to upgrade to high-efficiency cooling infrastructure could be considered ahead of repairing or replacing critical components to avoid downtime from failures. While a move to on-chip, on-rack, or different physical cooling architectures can be expensive and come with new operating procedures, benefits can accrue for example from increased compute density for AI workloads.
For new AI-ready data centres, the options are more intensive and can open up a fuller range of cost/risk/value assessments of available and emerging technologies. Some data centres, or their corporate centres of technology excellence, already continually assess and even trial new technologies in order to keep up with the latest compute densities, space utilisation and operating efficiencies.
The corporate weather risk may also drive new investment to enhance redundancies in power supply and cooling systems, and geographically disperse backup facilities to minimise downtime risks.
Overall, regularly assessing data centres for vulnerabilities to extreme temperatures and implementing improvements can be crucial for stability. For data centres to adapt to long-term changing climates, a collaborative effort is needed between stakeholders within and around each facility. Data centre operators, government agencies, and technology providers can work together to develop comprehensive adaptation plans.