Keysource Blog

The Here and Now of AI in the Data Centre Industry

Apr 5, 2019 11:40:33 AM

Discussions about the use of artificial intelligence (AI) in data centres tend to focus on the art of the possible, rather than the power of the actual – that can make a real difference today. Richard Clifford, head of innovation at Keysource, explains how those in the industry should be using technology to deliver practical solutions to routine problems and industry challenges.

The complex nature of data centres means that, even when operating at part-capacity, there are thousands of moving parts which can fail and lead to serious implications. Building in resilience to the mechanical, electrical and plumbing (MEP) design can be a significant step in reducing the risk. However, it cannot remove it totally, and facilities’ operations can look very different once they’re built.

As the central point for most organisation’s IT operations, their upkeep is vital. Take the failure of the data centre powering British Airways’ systems in 2017, as an example. Attributed to human error, an engineer disconnected the power and failed to follow the correct procedure to reinstate it. The fault left thousands of passengers stranded, cost the airline almost £60 million to fix and led to a legal battle in the High Court.

The importance of keeping these critical environments up-and-running is evident and the industry has long spoken about how technology can be used to achieve this. One of the most discussed solutions is AI, which many within the industry have long championed as the best bet for implementing efficiencies and savings. The more exciting, broader aspects of these discussions, from its potential for using robotics to its energy saving capabilities have made those in the industry sit up and take note. However, AI can also offer more simple benefits to tackle industry challenges in the present day and these are often overlooked.

AI in action

Despite the apparent benefits of AI, and the technology’s predicted $15 trillion boost to the world economy by 2030, there have been plenty of murmurings about some of the detrimental impacts AI could have on jobs. It’s this, in part, which is holding many back from deploying the technology and reaping the rewards. In many ways, AI should be viewed as a gift to the industry rather than something that is going to take away jobs tomorrow.

And when you shift the discussion away from theoretical pipedreams to practical applications, it’s easier to see the immediate gains.

One way that AI can be applied to estates now is through the use of sensors to monitor and track mechanical, electrical and hardware systems, allowing engineers a much earlier insight into when things may be going wrong.

This can also help us take some of the guess work out of maintenance, which can lead to issues being addressed too late and time spent fixing others too early.

Taking constant readings of data across multiple sensors can then feed into an AI system, which analyses patterns and helps spot where small fluctuations in vibration, sound or temperature might create much larger problems further down the line. Often this application gets confused with Building Management Systems (BMS), which many estates managers, engineers and property professionals already use. The difference is that where a BMS may alert you when a system has failed, AI can spot the early symptoms instead.

There are two benefits here. Firstly, the AI can process data far more quickly and more accurately than humans. Those in charge of maintenance can glean previously unavailable information on their infrastructure as soon as they deploy these AI systems. However, they really come into their own when the algorithms that sit behind them start to self-learn.

Each time a problem is diagnosed by AI and then solved by an engineer, this is fed back into the AI’s log. Over time, multiple instances of the same issue will allow the AI to develop an understanding of common malfunctions and what causes them – offering suggestions to engineers long before anyone has had to move from a desk. As it continues to learn, the AI will eventually alert engineers to the potential future failure of an entire cooling system, for example, before the human eye would spot any issue in a routine system check. This, in turn, will allow engineers to diagnose problems quicker.

Secondly, it can help ensure that the maintenance team can send out the best engineer to solve the issue. AI can determine when an ostensible mechanical fault is actually a problem with an electrical process elsewhere.

Both of these benefits in tandem foster a ‘stitch in time’ approach to maintaining estates.

At the cutting edge

These aren’t the only benefits and challenges that AI offers the industry. The emergence of the internet of things, smart cities and connected vehicles are creating new challenges around the need to access data at speed.

For an autonomous vehicle to function, for example, data needs to be instantly available to it. One data centre in the south of England can’t operate autonomous vehicles in Scotland, for instance, since the physical distance causes a lag between the data centre and the car.

This is partly why data centre owners and operators are looking to deploy multiple, smaller data centres spread across a wider geography. The industry often calls this “edge-of-network computing” or “edge” for short.

To return to the connected vehicle example, as a car travels it’s constantly connecting to its nearest data centre, which uses the analytic data at the edge to improve vehicle performance, reliability and driver experience. It also ensures vehicles are kept up-to-date with the latest software and equipped for autonomous or semi-autonomous operation. Being connected to the nearest data centre is vital – an extra millisecond lag in the process could easily cause a driverless vehicle to fail, rendering the benefits of any technology pointless.

These interconnected facilities, strategically deployed around multiple locations, are going to be a huge factor in the success of the future connected world. However, there’s a key challenge in terms of maintenance. The simple fact is expert engineers cannot be everywhere at once, so a solution needs to be found to minimise the mean time to repair (MTTR) and support regional teams with expert knowledge.

The insight that AI-enabled monitoring systems can provide will therefore be key to delivering edge. The technology can monitor for potential electrical and cooling problems that are likely to arise, so an engineer can fix individual issues without the need to go to every site for routine check-ups. It also ensures that the correct jobs are matched to the correct skill level. Skilled engineers can use the real-time data from sensors to talk someone through how to fix minor issues rather than making a 50-mile trip themselves. This will prove invaluable in tackling the skills shortage faced by the industry by allowing them to spend more time dealing with bigger issues.

Looking ahead in the industry 

Naturally there is a place for future gazing in the tech sector. But, as ever, many in the industry are running before they can walk when it comes to AI – thinking of the theoretical before the practical applications.

Using AI to create efficiencies in maintenance might be less sexy than other proposed applications for the technology. But it has profound benefits for cost and labour efficiency, uptime and resilience – everything that we see clients setting KPIs for. Embedding it in the industry now will go a long way to preparing for the imminent challenges of a more connected, data-reliant UK.


Topics: Transport, data centres, Insight

Richard Clifford

Written by Richard Clifford

Richard has extensive experience in delivering strategic client engagements, from design to operation of Critical Environments. Championing Innovation, Richard is instrumental in driving value for our clients, evidenced by some of our most pioneering and award winning projects and solutions.