Artificial intelligence makes data centers look like science fiction movies, but it comes at a cost

Airplanes are already practically flying themselves and we are very close to having self-driving cars. This is all becoming a reality thanks to the wider application of machine learning and artificial intelligence. Will the same fate befall the data centers that underpin our e-societies? Will our digital infrastructure start to think for itself and become ‘self-sufficient’?

Toomas Kell explaining the function of AI in data centers.

If we were to be a little provocative, we could say yes: machine learning and artificial intelligence, in this article referred to as AI, has already become a major game changer.

Optimisation beyond human capabilities

The modern AI enables technical systems to sense their environment, process information, and solve problems. The computer collects data with the help of sensors and processes and responds to the data. By analysing the effects of their past actions and working autonomously, intelligent systems can adapt and evolve their behaviour. This means that they have the ability to learn. AI is undoubtedly at the heart of the digital transition, as it enables us to perceive, manage, and respond to significantly larger and more complex sets of information than humans at lightning speed.

Data centers, where the internet physically ‘lives’, in simple terms, are by their very nature complex and multi-layered systems of technical devices. With this in mind, it seems logical to apply AI, as there is plenty of room for optimisation. Of course, in this case, it has to be done without compromising on reliability.

The big players are also looking for a place for AI in their data centers

‘Large data center operators such as Meta or Google have been looking for and finding ways to make data centers more autonomous for a long time. On the one hand, this arises from the desire to introduce new technologies and, on the other hand, the wish to ensure the flawless operation of more computing power and a longer lifespan of the equipment,’ says Toomas Kell, who is responsible for the daily operations of the most modern data center in Estonia, Greenergy Data Centers. The application of AI is therefore a much broader trend.

‘In data centers, AI can, for example, monitor and predict electricity consumption, estimate the need for cooling in the different areas of server rooms and adjust the cool air flows, control the removal of heat from liquid-cooled equipment, estimate the lifetime of equipment depending on its usage ... The list goes on,’ explains Kell. Ultimately, the outcome of the work of smart machines is measurable in terms of energy efficiency and hours worked, or money.

AI improves the lifetime of equipment as well as security

Going into further detail, AI collects feedback from the premises and equipment of the data center and adjusts the operations of the data center based on this information. ‘For example, this has freed our staff from the need to perform regular checks of environmental indicators in the server rooms and recalibrating them,’ explains Kell. Even in the control center, you do not have to keep an eye on every single indicator all the time.

AI extends the lifetime of equipment by ensuring optimal operating rhythms. Thanks to this, the equipment can be used longer before replacement, which saves money. AI also optimises maintenance cycles: ‘For example, if a piece of equipment is otherwise scheduled for annual maintenance, AI will calculate when it would be most optimal to perform the work based on the load on the equipment and an analysis of the alarms, faults, etc. which have occurred in the course of the functioning of the equipment. It may be in a real need for maintenance in three years instead of a year,’ explains Kell.

Help with security

There are also benefits in terms of security. Moving around the data center building, you have to pass through various gateways, which can be accessed by using biometrics or access cards. At a point of entry, AI monitors via video whether the number of people who have checked in matches the number of people actually in the room. If not, no one can move further and the control center will receive a signal regarding suspicious movement. The smart machine also helps to keep an eye on security cameras in data centers and spot threats instantly. Let us consider how many people it would take to monitor, say, 200 or 400 video feeds in parallel?

Furthermore, AI can help to calculate, for example, when it would be necessary to increase the power capacity due to a higher load from the IT equipment of a certain client, or find the most appropriate location for a new client in the building. This, in turn, creates an input for the most effective use of the capabilities of the data center and for construction or development work.

‘All of this does not mean, however, that people no longer have to come to work. Far from it. AI needs to be carefully monitored to make sure that its decisions are correct. People do look at the bigger picture, though, instead of a constant stream of smaller pieces of data,’ says Kell. So it is not true that AI is taking jobs away from people, but a person using AI is actually taking jobs away from a person not using AI.

The four risks

All good things have a downside, though. Risk mitigation is required also in the use of AI.

The first risk lies in setting it up properly. It is essential that humans have the ability to check the input data and algorithms which are used by the AI to make decisions. ‘To mitigate this risk, each organisation can put in place its own processes for how and when these decisions are reviewed,’ Kell says.

Second, the level of discretion of the AI must be defined. ‘This is where the clear benefits must be compared to the potential risks. AI can also simply offer people different options to choose from. There must certainly be a way to stop the functioning of artificial intelligence,’ he adds.

Third, the use of AI must not lead to learned helplessness. Every organisation must retain the capacity to understand and analyse processes and their interrelationships beyond the computer. ‘The use of artificial intelligence does not mean that smart and highly paid people can be made redundant. Rather, AI should empower them,’ says Kell.

Fourth, the interaction between different pieces of smart software needs to be defined in detail. As AI is used for a wide variety of tasks, it may happen that several systems with different operating principles are operating at the same time. For example, solutions created by different manufacturers may function in different ways and combining them may be challenging. One option is to prefer choosing systems from the same manufacturer.

‘At Greenergy Data Centers, we chose Siemens,’ says Kell. ‘The fact that the company has access to continuous technical support in Estonia, which is one of the most important components of operating the system, played an important role.’

The future is looking increasingly like a science fiction movie

If you look at the huge players such as Google, Facebook, and a few others, they are occasionally moving to fully automated autonomous data centers.

However, it should be kept in mind that such data centers serve only one company, which opens the door for extensive standardisation. An institution serving multiple customers has to take into account different interests and needs, and the benefits may not outweigh the risks.

First and foremost, AI in a data center means efficiency

All in all, AI has already found its place in data centers. Thoughtful implementation results in the following:

• higher energy efficiency, which will save electricity and money. It will also help to meet the increasingly stringent environmental requirements;

• less unskilled work, motivating employees to learn and contribute at a higher level;

• predicting bottlenecks in systems before they become issues, thereby increasing reliability;

• greater security, because machines can do much more in any moment in time;

• extending the lifetime of equipment through needs-based maintenance, thereby avoiding unjustified costs and consumption.

Therefore, the implementation of AI is more of a growing trend, helping to balance the footprint of the IT world on both the environment and our wallets.