Artificial Intelligence (AI) has been causing a stir across nearly every industry vertical, heralding the next great technological shift akin to the dawn of the computer or smartphone. While the term goes as far back as 1956, coined during initial work involving computers mimicking base-level human reasoning, skyrocketing data demands and advanced operational frameworks have helped AI become increasingly germane to a growing number of needs. As a result, AI promises to revolutionize the way we maintain our critical data centers. Predictive maintenance is now the most successful component of a maintenance program in terms of achieving the highest equipment effectivenessOpens a new window level, and is a particularly valuable tool for battling the data center’s biggest adversary: downtime.
However, as a relatively new application in the data center, there aren’t many footsteps to follow. Additionally, due to a lack of overlapping domain knowledge between AI and data centers, a data center team navigating AI installation can seem like asking a mechanical engineer designing car engines to lead the design of SpaceX’s Falcon 9 rocket. Obviously, this can make it challenging for businesses to explore new AI applications while simultaneously maintaining top-tier standard business performance. Luckily, there are teams in the data center sphere that have found ways to successfully strike a balance.
ROOT Data Center has completed the initial phases of their AI project, which is the world’s first instance of AI used in a colocation data center to reduce customer downtime risk. ROOT has developed a five-year strategy that will eventually lead to a widespread utilization of AI sensors, enabling their facility to automatically monitor and predict potential failures. Their plan began with installing and deploying sensors within the generator platform of a 5MW data hall. Since installation, these sensors have participated in over 3,000 training sessions with the generators, representing 250 hours of monitoring augmentation, to learn baseline operating levels and recognize irregular indices. During this time, the team was also responsible for building and deploying additional data halls, including the beginning stages of a new purpose built 10MW data center in Montreal.
At the top of the list of priorities for this mission was its replicability. At its heart, this project was designed to ensure reduced downtime risk for data centers while being cost effective and efficient so other data centers around the world can implement a similar program. To create a framework for other teams’ mobilization with minimized business disruption, the first step is to understand the drive for AI and how implementing it will wholly aid to the growth of the data center industry. For ROOT, the motivation for an AI deployment grew from the company’s central goal of being an industry disruptor and spearheading innovation to drive success for all. Especially in the age of the cloud and the edge, demands on the data center are growing every day. As one of the largest and fastest growing cloud companies, ROOT pinpointed AI as a critical step towards facilitating a transformation to support its operational methodologies and handle sky-high demands. Keeping eyes on the prize and understanding what can be achieved for data centers across the world will inspire data center teams to push forward even when roadblocks pop up.
Designing, operating and constructing critical infrastructure is no easy task, and it demands dynamic responses to a range of issues. Building a team of individuals that are driven by a strong desire to learn and overcome is essential to AI implementation success. Problem-solving was one of the critical skills at which the ROOT team excelled. First order problem solving allows a team to anticipate their project path with the highest probability of a successful outcome. Applying this tireless work ethic and passion for critical thinking to a clear framework, in ROOT’s case a clear four-part project where the AI was implemented in a stepwise fashion, was the key. This allowed the team the ability to simultaneously deploy 15MW of new capacity, break ground on a greenfield 10MW building and enhance ROOT’s online learning management system to rapidly onboard new members and more.
This step-by-step approach brings up another key point in achieving AI implementation while maintaining standard business practices: keep it gradual. The old saying ‘Rome wasn’t built in a day’ comes to mind here. The way ROOT approached it was in steps; each segment of the project will last around a few years to ensure each step is successful and has enough testing, then it will be built upon. First, the team simply deployed the sensors, collected data and achieved baseline operating understanding. Next, the AI will be expanded into the primary monitoring systems, then from there into augmenting controls with help from operators. In 2022, the team is finally looking to move into an AI-first operating system with human fail safes.
AI presents huge and revolutionary opportunities for battling even the most challenging aspects of data center operation. Admittedly, implementation can seem like an uphill battle, especially when trying to manage an AI team and avoid business disruption. Luckily, there are teams that have done it, and their stories are here to light your way. ROOT successfully reduced downtime risk while bolstering operational efficiency, all by implementing a clear, gradual process and maintaining a dynamic approach to roadblocks.