How will AI disrupt data center operations?

ROOT Data Center are a Montréal-based company and the first in its sector to implement AI to support their essential human workforce while solving a critical need

We recently spoke to their CEO AJ Byers about the company’s vision and use of AI and its impact on its business operations, workforce and growth.

In your opinion, how will AI disrupt data center operations?

When it comes to data center operations, we’re starting to see that AI is less disruptive and more supportive. Our approach to AI technology adoption was to enhance our existing operations and further reduce risk of downtime. Other implementations that we have seen, notably in Google data centers, has been to increase power efficiency, yet another example of taking an existing operation and making it better.

You’re one of the first collocation players to bring AI into your DC operations. What motivated you to take on the task?

We are, to the best of our knowledge, the first data center to implement AI in order to reduce downtime risk. Since taking this approach, there have been other players in the data center space that have followed our lead, implementing LitBit’s product for similar purposes.

We are motivated by challenging convention in data center design and operations in order to provide our customers with the best possible data center experience. In our implementation of AI, we are motivated to do everything we can to reduce downtime risk for our customers and their critical infrastructure.

Does this mean downtime will be a thing of the past soon?

Reducing downtime to zero is an immense challenge. Design standards that increase power redundancy are doing a good job, but are also very complex and can become hugely expensive. Theoretically, a data center could have dozens of back-up generators for each power line-up, but the costs would be incredibly high for the operator and its customers. Designs that are more practical and efficient are important, but there is always a small amount of risk. Recognizing that need, we are utilizing AI to reduce risk of downtime without enormous cost.

How important a component were LitBit in delivering this successfully?

They were integral to this process – LitBit provided the artificial intelligence capabilities as well as the remote sensors.

So how does it work?

LitBit’s solution includes a “persona” of artificial intelligence with a domain of expertise in the data center space. In order to gather information, the system uses a set of remote eyes and ears that are placed by ROOT technicians throughout our data center. ROOT technicians provide additional learning to the AI persona by “teaching” them about our operations while conducting preventative maintenance. Much like a skilled auto mechanic might know by sound when a fan belt will soon need to be replaced, the persona learns to predict the need for preventative maintenance. This learning capability is combined with fine-tuned sensors that capture sound beyond human capability, as well as the ability to be in more places at once with multiple sensors.

What are the implications for the IT Ops folks?

The ROOT IT Ops team are supported in having more sensors covering more area of our operation, 24 hours a day, 7 days a week. They are then free to do different work to serve our customers. Our customer IT operations gain increased confidence that our downtime risk has been reduced.

It’s not just AI that makes ROOT different. Tell how you’re helping your clients reduce their carbon footprint?

There are three ways that we are reducing our carbon footprint – which is an increasingly important conversation for energy-hungry data centers. First, strategic site selection led us to place our operations in Montreal, Canada. The Hydro Quebec grid here is 100% renewable with almost all of that power coming from hydro-electric generation, which has an incredibly low carbon footprint. It also has a very low consumption of water, as some other generation methods like coal and nuclear take significant amounts of water out of the drinking system. Second, that site selection in Montreal provides us with a naturally cool climate which does not necessitate as much cooling as other locations in warmer climates.  More of our power consumption goes to compute, rather than cooling.

Lastly, we chose to use Kyoto Cooling, which is a free-air heat exchange system that leverages the climate to cool our data halls.  This is an incredibly efficient system that allows us to use free cooling 90% of the time, rarely having to rely on CRAC units.  This approach also reduces our water usage as compared to popular liquid cooling methods. All these systems combined have enabled us to achieve power usage efficiency (PUE) of 1.17.  For reference, PUE is a ratio of how much energy coming into our facility is used for IT equipment rather than facility power. To put our number in perspective, Google’s facility is a benchmark at 1.12 and the industry is closer to 1.3.