Is AI the key to going green?

Is AI the key to going green?

Jim Chappell, global head of AI and Advanced Analytics at AVEVA, shares how data centres can exploit artificial intelligence to become more efficient, sustainable and robust.

All companies, large and small, are under increasing pressure from investors and other key stakeholders to minimise the negative impact their operations have on the environment. The data centre industry is no exception.  

Data centres consume a vast amount of electricity, roughly 3% of all electricity generated on the planet. This accounts for approximately 2% of global greenhouse emissions, which is the same as the aviation sector. This is only set to increase as the world embraces the Fourth Industrial Revolution and the Internet of Things, with an increasing number of objects connected to the internet and relying on data centres. 

IDC predicts that global data usage will increase from 33 zettabytes in 2018 to 175 zettabytes by 2025 – an increase of 430%. Furthermore, the more powerful the data centre is, the more cooling it requires. These cooling systems also consume a significant amount of energy which has led some operators to move data centres to colder climates or find less energy-intensive cooling techniques. Microsoft has even explored submerging a data centre underwater. 

As demand on data centres increases, it is clear the industry needs to design facilities for maximum energy efficiency and minimum environmental impact. The operation of a ‘green’ data centre must also consider the need for its IT technology to use less energy than is needed to cool it.  

How can AI be harnessed to better sustainability?  

A data centre, powered by artificial intelligence (AI), is nothing new. For some time now, data centre operators have been aware of the significant operational benefits of deploying AI in their facilities. AI can allow data centres to operate autonomously by automating the routine tasks involved in the maintenance and monitoring of these centres.  

A common solution is predictive analytics in the form of machine learning. AI can identify anomalies in processes or equipment, which indicate performance issues or a deterioration in an asset’s health, well in advance. Sophisticated AI can even identify the probable root cause of the problem and recommend a course of action to best remedy and optimise a given situation. Issues can be identified and corrected quickly and accurately before they have an impact on operations.  

AI can also significantly reduce costs by minimising downtime and increasing output. A McKinsey study estimates that predictive maintenance enhanced by AI can reduce overall maintenance costs by as much as 10%, downtime by 20% and reduce inspection costs by 25%. With the price of storage and computing power plummeting, reducing operational and energy costs can turn a break-even data centre into a profitable operation. Indeed, Gartner predicts that data centres that fail to deploy AI effectively will become economically and operationally defunct. 

Beyond productivity and profitability 

Discussions around the benefits of using AI in data centres often focus on how it can increase productivity and profitability, but neglect to talk about how AI can make a data centre more sustainable. 

Using historical data collected from smart sensors, such as data output, temperature and humidity levels, AI can train deep neural networks to optimise the performance of a data centre and make it more energy efficient. Moreover, prognostic AI can forecast future events, such as surges in demand or temperature changes, and adapt the system variables accordingly. This not only prevents the data centre from going beyond its operating constraints, but also ensures it operates as efficiently as possible. 

Less energy-intensive data centres require less cooling, which in turn reduces total energy usage. AI is already being deployed by some of the biggest players in the industry. Google used DeepMind machine learning in its data centres to directly control the cooling systems, which resulted in a notable 40% reduction in energy consumed. 

Not only does the implementation of AI constitute a significant cost saving, but more importantly it dramatically reduces harmful emissions and the carbon footprint of companies who rely on data centres. A recent PwC study found that deploying AI across business operations could reduce global greenhouse gas emissions by as much as 4% by 2030 – the equivalent to the combined 2030 annual emissions of Australia, Canada and Japan. 

Where should data centres start 

To get maximum value from AI, businesses with deployed data centres should first look at their IT and control infrastructures. If they are collecting data from their control systems and/or energy management systems, then they are excellent candidates to benefit from AI and reduce overall energy consumption. And as they add additional sensors to their infrastructures, AI can provide increased value and sophistication to achieve an even higher level of efficiency. 

Often, some of the first benefits gained from the implementation of AI in data centres include detecting equipment that waste electricity. AI can quickly identify underperforming assets, as well as those with maintenance issues; both of which result in the consumption of excessive power. Across a large data centre, the wasted electricity from these types of assets can quickly add up to a significant cost and a significant impact on the environment. This is fundamental and a recommended first step in AI implementation. 

Cooling is absolutely essential for data centres, and temperature hot spots can occur when least expected. As these situations worsen, computer equipment can start to fail, strange anomalies appear, and energy is wasted. Further, as equipment heats up, it can become less efficient, requiring more energy to run. This can result in a bit of an efficiency degradation spiral. Since hot spots often slowly get worse over time, they can also go undetected for quite some time, resulting in potentially serious issues. Automated monitoring with AI analytics is a ‘best practices’ method of early detection of hot spots, resulting in increased operational efficiency and overall energy savings. 

Enhanced AI capabilities are continually developing and evolving to minimise energy consumption, minimise downtime, and maximise efficiency. Some of the newer areas where AI is helping have to do with balancing load across servers as well as within a given server across multiple CPUs in order to minimise overall heat generation and, thus, power consumption. Power distribution at a data centre can also be optimised through AI.   

Associated anomalies detected through aberrations of multi-variate patterns, as well as an overall analysis of situational awareness related to power delivery is an area that is continuing to evolve. Additionally, sensors are becoming more prolific, providing AI with additional ‘raw material’ to perform further analysis and provide more sophisticated insight so that data centres can continue to expand while using less energy. 

There is no doubt that sustainability has become a top priority for business executives, investors and governments alike. Data centres are, and will continue to be, an integral part of the data-driven economy we live in; therefore, finding ways to reduce their contribution to carbon emissions is critical. This is where AI can play a major role. By optimising operations and increasing energy efficiency, AI can ensure that data centres become more sustainable as the world continues its ascent towards a greener global economy.