Many data center operators are today faced with tough decisions over which power source their facility should rely on.

This has become particularly acute in Europe due to the looming danger of blackouts in countries across the continent. Indeed, high prices for essential goods or commodities can have unexpected knock-on effects.

For example, households across Europe rely on gas for heating homes, but with the price of gas going through the roof and the risk of shortages, many are considering alternatives this winter. The main alternative to gas heating, of course, is electric. But if everyone turns their electric heating on at the same time in response to the same cold snap, rather than using gas, there’s a high chance of brownouts. Power grids across Europe may no longer be reliable.

The same problem arises in the gas market because of the current shortages and the resulting spike in gas prices, which also affects power prices and, hence, the business models of data centers.

Although it is common practice by governments to maintain sufficient gas storage to overcome any short-term supply issues, this cannot guarantee long-term gas supply. In the UK, for example, gas storage is barely sufficient for even a week’s demand.

The question is, if a gas and power shortage leads to rationing by the government, will data centers count as critical infrastructure or be forced to take emergency action?

A third source of power is renewables, of course, which can contribute to data center needs, but cannot guarantee continuity of supply. This is obvious for solar and wind, but even the supply from hydropower plants could become uncertain due to regional droughts. The driver for renewables is a combination of price and marketing, as it strengthens an organization’s green image; and, to be clear, we do not mean green certificates, which could better be described as ‘green washing’.

So what can be done?

The only guaranteed fall-back strategy for data center operators is backup-generators.

TÜViT has been evaluating and certifying data centers according to their availability and physical security for 20 years. Our TSI.STANDARD criteria catalog, as well as the EN 50600-2-2 standard, define back-up generation as the most appropriate measure to control outages and shortages, and to provide genuine power independence for data centers.

None of the standards specifies the fuel for the gensets, though. While the majority run on diesel, natural gas-based backup has gained a significant traction in recent years. Nevertheless, organizations using diesel for backup typically keep storage onsite, whereas gas gensets usually rely on continuous gas delivery, which might be caught-up in any become a problem in the future.

There is also another aspect that has to be taken into account, which is the layout of the generators and their usage scenarios. The International Organization for Standardization (ISO) provides with the ISO-8528-1:2018 standard definitions of on-site power production and divides generator set classifications into four rating categories:

• Continuous operating power (COP);
• Prime-rated power (PRP);
• Limited-time prime (LTP);
• Emergency standby rating (ESP).

We have seen at many small and mid-range data centers that power backup is taken for granted, and that their generators will only provide power for between no more than 200-to-500 hours per year – 20 days per year, maximum. Their backup infrastructure is designed to operate as LTP or as ESP generators. They can mitigate the effects of grid outages as long as the outages only last a few hours. If the supply from the grid is missing for days, or even weeks, the COP or PRP-types come into focus.

COP-rated generators provide power in a continuous, constant load with an output of up to 100 percent for an unlimited number of hours per year, but with no overload capability. PRP-rated generators, meanwhile, provide power with an indefinite running time in a variable load setting – average load should not exceed 70 percent. PRP and COP are similar in terms of providing an ongoing, steady power supply, but PRP can run in both variable and overload capacities, which COP cannot.

Within the next five years, some hyperscale data centers (>100 MW) plan to eliminate back-up generators entirely in order to save money and to circumvent environmental obligations regarding diesel backup, but this is not a good idea in terms of guaranteeing reliability.

The high-availability property of a data center relies, increasingly, on the power sources used and their reliability and availability. Operators are advised to use a mix of power sources simultaneously, and to have additional back-up generators capable of handling the challenges of providing continuous, constant load.