The moment a data center comes to life, everything has to work without exception: if one part fails, nothing works. That’s the everyday reality of Goran Vujević, a commissioning manager who leads the process of bringing data centers into operation - known as commissioning.

For the past year, he has been living in Winschoten, a small town in the north of the Netherlands near Groningen, where he is working on one of the more complex projects of his career.

The project he is currently working on involves collaboration with major suppliers such as Vertiv, as well as an end client from the hyperscaler category, like Google. It is precisely the scale of the project that defines its complexity.

“We’re responsible for making sure everything is under control,” says Goran, quickly checking messages on his phone before our conversation dives into the details.

What does it actually mean to “put a data center into operation”?

The job of a commissioning manager begins once everything has been installed, and their role is to coordinate and oversee. He participates in team meetings, assigns tasks, and connects teams coming from different companies and disciplines - from equipment manufacturers, electricians, and testers to QA teams and BIM engineers.

“Others report a problem, and we ensure a solution.” In practice, this means that the commissioning team stands between all involved parties and the ultimate goal—a fully functional data center.

The commissioning process takes place in multiple phases and is clearly structured,” Goran emphasizes, adding that the verification process begins already at the factory, where initial equipment tests are carried out, and continues on-site through several levels of checks, from L1 to L5 phases. Each phase has its own checklist and must be documented. After the initial testing, Integrated System Testing (IST) follows, where all components are tested together for the first time. Between phases, additional checks—so-called walkthrough tests—are also carried out to confirm that the system is functioning after each intervention. Only once all phases are successfully completed can the data center transition into an operational state.

The project Goran is working on has a capacity of 66 megawatts, divided into six units of 11 megawatts each. Each of these units goes through its own commissioning cycle, and one such cycle can take years - “For example, one part of the project started in 2024 and has still not completed all phases.”

In addition to the complexity of the technology, the reason for this is also the number of people and teams involved in the process. A single project can involve more than 20 different companies, each of which must deliver its part and leave a clear trace in the documentation. Although the large number of people involved in such a process can be challenging in terms of communication and continuity, the common goal is simple - a fully functional system in the hands of the client.

With hyperscalers, there’s no room for improvisation

Clients like Google fall into the hyperscaler category, which means they operate across a large number of data centers simultaneously while adhering to strictly defined process standards. In such an environment, there is little room for improvisation. “On smaller projects, there are various challenges we encounter, but here we are talking about structured processes, and every phase, test, or modification goes through clearly defined protocols.”

Data centers of this level must meet Tier 3 standards. In other words, such a data center must be operational 99.982% of the time throughout the year, as downtime can be extremely costly. “This means we can afford only 96 minutes of downtime per year,” Goran emphasized. 

That is why infrastructure reliability is a business necessity in projects like this, since any interruption means starting over, and the scale of the system is best understood through the numbers. A single graphics card can cost up to 50,000 euros, and a server contains ten such units, while a rack holds eight servers. All of this requires precise planning of power, cooling, and connectivity.

An emerging industry and a mindset that sets us apart

Goran has been working at Barrage for eight years and remembers how the team he is a part of started with just five people, while today it numbers around 70 employees. The reason is clear - choosing a business niche and the extremely high demand for data centers, driven by the development of artificial intelligence, has resulted in a large number of projects that Barrage is working on. “We are in a situation where we sometimes have to turn down projects, and we are always looking for people to work in the field.”

And finally, in this third edition of Bytes, I’m asking the same question again: what sets Barrage apart from other companies in the field?

“We always learn through projects. Sometimes you’re not 100% sure, but you have to take responsibility. Barrage teams are often called to projects to solve problems when other teams aren’t available, for example, on weekends. But the difference lies in the approach to work and the way of thinking - what some would call a ‘mindset.’ We treat every project as if it were our own,” Goran concludes.

This approach quickly becomes evident on-site. Although they were less experienced at the beginning, today other teams seek them out for advice. Because the moment a data center is “switched on” - there is no second chance.