How artificial intelligence makes a data center more efficient

Can artificial intelligence play a role in optimally setting up a modern data center? Martin Matse from Perf-iT thinks so. He states that AI can be an important tool to perform Data Center Infrastructure Management (DCIM). During the IT Infrastructure event Matse dives deeper into the benefits of applying artificial intelligence.

By: Dimitri Reijerman

At a time when energy consumption in data centers is increasing rapidly, optimization is a must. But it is not a given that seemingly obvious technical adjustments actually benefit a data center. Matse: “Data centers must continuously optimize. A magic value is the PUE (Power Usage Effectiveness). But how exactly are you going to do that? And what is the consequence of your optimization? For example, you can lower the cooling water temperature by two degrees, because then you may save money. But what is the impact of this on the data floor? That is difficult for a person to reason.”

Artificial intelligence, a technology that is developing rapidly and has more and more applications, can therefore be a useful tool to improve the availability and efficiency of a data center. Matse says: “What we do, together with TNO, Vortech and Actiflow, among others, is create a digital twin. This is a digital copy of reality that also incorporates all physical properties of a data center, such as air flow and temperatures. The great thing about such a digital twin is that you can test different scenarios based on data. Based on this new data, the AI model will tell you exactly what the impact is on the data floor.”

Sensors & algorithms

Setting up a well-functioning model requires some preparation, says Matse: “What we are doing now: we make a basic model and install sensors in various places. Based on the input, the model has a certain expectation. And based on the sensors, this model sees every time what the actual situation is on the data floor. The algorithms become increasingly accurate: the model continuously adapts to the newly found reality.”

Applying artificial intelligence for DCIM is still relatively new. That is why many tests are still being done, says Matse: “TNO has developed and patented the basic technology. We have now tested it in practice at two data centers: at Infrabel - the Belgian ProRail - and at NL-DC. Applying this DCIM technology delivers a very fast ROI. And things happen on the data floor that were not expected in advance.”

To keep improving the model, many operating hours are needed: “A model is dynamic,” says Matse. “Based on the process and by placing more and more sensors, you see that the model is becoming more accurate. The best-known forms of AI, deep learning and machine learning, cannot make predictions outside of the data it has once seen. But a digital twin goes further, because you have also recorded physical processes in algorithms.”

There are currently few companies that offer DCIM based on machine learning and deep learning on the market or apply it in their data centers. According to Matse, this is due to the aforementioned limitations of these implementations of artificial intelligence. He gives an example: “In the US they have used deep learning to select applicants. They fed the AI with countless resumes of ideal candidates. Then all the women were rejected. How did that happen? The data provided did not include any resumes from women. So the completeness of your data is important. With our model this is less important because you actually use algorithms as the basis of your model.”

Future of DCIM and AI

In the future, the application of AI for DCIM applications will still encounter some challenges: “The challenge is mainly acceptance. For many people it is still very scary to have a computer model determine how a data center can be optimally controlled. But it is becoming more and more common. We will often first see an intermediate step: the model provides advice and does not yet actively manage the data center itself.”

Finally, Matse also points out the social benefits of this technology, especially in the field of sustainability: “I think there is still a lot to be achieved in the field of energy reduction within data centers. That is an issue throughout the country and especially in Amsterdam and the surrounding area. A nice by-catch of DCIM is the energy transition, for example for the reuse of heat from data centers. If you cool smartly, the return air temperature will also increase. This increases the efficiency of your heat pumps. And if municipalities become very strict about enforcing the use of residual heat, for example for the construction of a new district in Amsterdam West, then it is almost impossible to avoid using residual heat from data centers.”

Would you like to visit the IT Infra event free of charge and attend this lecture? Register now.