It's time to rethink our generator addiction

The data centre industry is currently facing a generator problem. The status quo for new build data centres is to provide the full site load with backup generator capacity without first considering whether it is the appropriate solution for the site. The proliferation of repeatable template designs, which allow for a consistent and scalable deployment, further stifles the ability to apply first principles to optimise the design site-by-site.

The Uptime Institute, an unbiased advisory organisation focused on improving business-critical infrastructure performance, efficiency, and reliability, states that on-site power production should be considered the primary power source, and local utility power as an economical alternative only.

Because of this requirement, many end-customers demand full generator backup and refuse to consider sites that cannot provide this, fuelling the existing misconception that it i’s a necessity.

This is not only an economic issue, but also a sustainable one. To achieve the industry’s net zero carbon goals, we need to change how we view the role of generators. We must embrace other options that are better for everyone. We should consider if complete, or any, generator backup is required on an individual project basis.

Re-evaluating their need for data centres

As data centres grow, with hyperscale campuses designed to support artificial intelligence and machine learning, sites connect to the grid at higher voltage levels, which significantly shifts the reliability dynamics. It’s crucial to distinguish between the reliability of distribution networks, which typically operate at voltages between 11-33kV, and transmission networks, which typically operate at voltages of 132kV and higher.

Data centre developers and their customers must adopt a more analytical approach, evaluating system reliability and assessing the historical performance of upstream networks and substations. This comprehensive assessment is crucial in determining the need for total generator backup capacity.

The industry standard of ‘5 9s’ availability, meaning 99.999% uptime, is an appropriate benchmark to manage risk due to failures on a system. However, when sites connect at the transmission grid level, including generators, in many cases, it often yields marginal benefits, and many of these campus developments may achieve ‘5 9s’ without any generators. Additionally, the incorporation of generators introduces significant costs and complexities, as well as environmental consequences. The ecological toll of this conventional approach is underscored by embodied carbon from equipment production and operational emissions during routine testing.

At Cundall, we’ve seen the disparity between perceived necessity and practical application first-hand. For instance, data centres with over 50 idle generators for a decade have never been run outside of testing and maintenance. And yet, we’ve also seen a select few sites with highly resilient utility connections highlight the viability of alternative approaches, achieving ‘5 9s’ availability without generators. It is possible.

In regions like the UK, the Energy Networks Association Engineering Recommendation P2 Security of Supply mandates that for consumer groups of 60MW-300MW, the immediate restoration of power within a minute of the first circuit outage and three hours for the second circuit outage.

So, what needs to change?

As an industry, there are a few changes we can begin to start implementing to make a better alternative a reality. The first step is to dispel the misconception that backup generators are an absolute necessity. Many tenants believe they need backup generators and include this in their mandatory leasing requirements. This perception makes backup generators a requirement for the industry, and breaking down this barrier is the first step towards true change. To achieve this, end customers should take the time to review the sites’ grid and data centre reliability in detail, review how this relates to their needs and find a site that is a better fit for them economically and sustainably.

The next step is to change the philosophy of resilience. Currently, when reviewing a site for a data centre, customers only look at the resiliency at a site level. A more holistic approach is necessary, where several discrete sites connected at different grid parts are considered. Viewing these systems from a broader scope and providing redundancy through the IT infrastructure will enable generator backup only for the vital network traffic, which can then be diverted to another data centre connected to the grid.

Some hyperscalers already incorporate this method to improve resilience, and if more organisations follow suit, we could end our dependence on generators alone.

Viewing electrical resiliency from the grid first rather than the site backup generators would also allow data centre operators to strategically choose their data centre locations close to major grid substations, ensuring resilience is provided through the grid.

Entering the reduction phase

Organisations often overlook real-life data that demonstrate the reliability of their data centres in favour of a perceived need for generators. They can find other sites and locations more suitable for their projects by analysing the existing data we’ve collected for decades. Doing so can result in significant cost savings and align more closely with net zero carbon goals. While generators will still be necessary in most projects, each should be evaluated individually to determine the best fit. We must break away from business as usual to meet the commitments of the COP21 Paris Agreement and our obligations. It’s time to reassess our reliance on generators and consider each site’s needs before installation.