Today’s data center is too big to "power" fail

With data centers increasingly vital to modern business, the cost of a power failure can be enormous. Protecting a company’s investment requires the latest technology.

For 75,000 passengers stranded around the world by a British Airways IT problem in May 2017, the cause of the disruption probably didn't matter much. They just wanted to reach their destination. But for the airline, it was vitally important to discover what lay behind the outage that resulted in the cancellation of 726 flights, a loss of $108m and untold damage to the brand’s reputation with customers and investors. The problem was traced to one of the airline’s data centers.

In August 2016, US carrier Delta Airlines had a similar problem. Its power outage, which took three days to resolve, led to the cancellation of 2,300 flights and cost $150m. In September 2016, an outage at the Global Switch 2 (GS2) data center in London lasted for less than a second but still took one customer offline – GS2 didn’t say what industry they were in – for two days.

Whether for processing customer information, running e-commerce sites or providing employees with access to cloud computing, data centers are vital to modern businesses. Almost every function of the digital world is dependent on them.

The cost of downtime is estimated at $9,000 per minute  and, as the GS2 example demonstrates, even a blink-and-you’ll-miss-it power problem can cause significant downtime. With the data-center market becoming ever more important to modern business and society, these risks will only be more acute in future.

Reliable backups are key

Data centers are power-hungry creatures. It’s estimated that worldwide they consume around 500 terawatt hours (TWh) per year – slightly more than the UK’s annual energy consumption and around two percent of worldwide greenhouse gas emissions, roughly equivalent to the emissions of the airline industry. And most of that energy doesn’t go on powering the servers; instead it’s needed for cooling – both of the servers and the power supply itself.

With all this in mind, it’s vital to have a reliable backup system in place to reduce the risk of outages. Power problems will happen, but a well-designed UPS will ensure seamless operation by drawing power from a battery until the backup generator is operating.

One example of changing technology can be found in the kind of batteries used in the UPS. Most of the industry still uses lead-acid batteries, which are big and heavy and need a lot of cooling. Accommodating them affects everything from cooling costs to the amount of space the data center needs – and even whether the floor should be reinforced to support the weight.

However, the latest lithium-ion (Li-ion) batteries offer greater power density, so they can be up to three-times more compact and six-times lighter. They also last longer than lead acid – up to 20 years, compared with four to six years. And because they work in higher temperatures, they require less cooling. That lowers costs and frees up more space in the data center for servers.

Li-ion is more flexible, too. The smallest lead-acid option provides five minutes of backup power until the generator takes over. For many applications that’s more than necessary and thus a waste of money. Li-ion can cover smaller gaps.

Batteries are getting smart

Another big benefit is that Li-ion batteries can be ‘smart’. It’s impossible to know when a lead-acid battery will fail. A data center might find out a failure has occurred only when backup power is needed, resulting in the kinds of problems that befell Delta and BA. Operators either have to accept that risk or invest in further redundancy.

Li-ion batteries, in contrast, are self-monitoring, so data-center operators are constantly aware of the battery’s health and don’t need to waste money replacing it too early. Saft’s Flex’ion Li-ion batteries are self-powered, meaning they can monitor their condition even without a power supply.

Of course, the battery is just one part of the UPS and the UPS is just one part of maintaining a reliable data center. However, as businesses rely on data centers more and more to support their digital operations, it’s vital that the underlying power system can be counted on in an emergency. The financial and reputational cost of failure is simply too high.