When the Power Goes Out: What Keeps a Data Center Online?
When we talk about data centers, we often think about servers, networks, cooling, UPS systems, generators and, today, AI.
All these technologies are important. But from my experience in data center operations, reliability also depends on the people, the procedures and the way the infrastructure is maintained.
A good example is a power failure.
When the main power is lost, the UPS systems immediately keep the critical IT equipment running. The generator then starts and takes over according to the site’s electrical system.
For someone outside the data center, nothing may seem to have happened. The servers are still running and the services remain available.
For the operations team, it is a very important moment. We need to know exactly which equipment is connected to which power source, which UPS is involved and which PDU supplies each rack.
This is why A/B power is so important.
Equipment with two power supplies should use two separate electrical paths. For example, Power A can be identified with black power cables and Power B with red cables.
One power supply is connected to Power A and the second to Power B.
The goal is not just to have two cables. The goal is to have two independent power paths. If both paths depend on the same equipment somewhere upstream, the redundancy may not be real.
Cable identification also helps during maintenance. If an engineer needs to work on Power A, it should be easy to identify the equipment still connected to Power B.
But colour coding is not enough. The power paths must also be clearly labelled, documented and checked.
The same principle applies to DCIM.
A DCIM system gives us important information about racks, equipment, power connections, network ports and capacity. But if we change something in the data center and do not update the DCIM, the information can quickly become wrong.
For me, one basic rule is simple:
If you change something physically, update the documentation.
This applies to equipment moves, PDU connections, network ports, fiber connections and rack layouts.
Cabling is also important. A clean and well-labelled installation makes maintenance and troubleshooting much easier. During an incident, knowing where a cable goes can save valuable time.
Finally, there is the human factor.
Data centers are becoming more automated, but when something unexpected happens, someone still needs to understand the situation, find the cause and make the right decision.
Technology provides redundancy.
Good operations turn redundancy into real resilience.
A reliable data center is not only built with technology. It is built with good procedures, accurate documentation, communication and experienced people.
Nabil Djaouablia
Data Center Engineer | Data Center Operations & Infrastructure
When the Power Goes Out: What Keeps a Data Center Online? When we talk about data centers, we often think about servers, networks, cooling, UPS systems, generators and, today, AI. All these technologies are important. But from my experience in data center operations, reliability also depends on the people, the procedures and the way the infrastructure is maintained. A good example is a power failure. When the main power is lost, the UPS systems immediately keep the critical IT equipment running. The generator then starts and takes over according to the site’s electrical system. For someone outside the data center, nothing may seem to have happened. The servers are still running and the services remain available. For the operations team, it is a very important moment. We need to know exactly which equipment is connected to which power source, which UPS is involved and which PDU supplies each rack. This is why A/B power is so important. Equipment with two power supplies should use two separate electrical paths. For example, Power A can be identified with black power cables and Power B with red cables. One power supply is connected to Power A and the second to Power B. The goal is not just to have two cables. The goal is to have two independent power paths. If both paths depend on the same equipment somewhere upstream, the redundancy may not be real. Cable identification also helps during maintenance. If an engineer needs to work on Power A, it should be easy to identify the equipment still connected to Power B. But colour coding is not enough. The power paths must also be clearly labelled, documented and checked. The same principle applies to DCIM. A DCIM system gives us important information about racks, equipment, power connections, network ports and capacity. But if we change something in the data center and do not update the DCIM, the information can quickly become wrong. For me, one basic rule is simple: If you change something physically, update the documentation. This applies to equipment moves, PDU connections, network ports, fiber connections and rack layouts. Cabling is also important. A clean and well-labelled installation makes maintenance and troubleshooting much easier. During an incident, knowing where a cable goes can save valuable time. Finally, there is the human factor. Data centers are becoming more automated, but when something unexpected happens, someone still needs to understand the situation, find the cause and make the right decision. Technology provides redundancy. Good operations turn redundancy into real resilience. A reliable data center is not only built with technology. It is built with good procedures, accurate documentation, communication and experienced people. Nabil Djaouablia Data Center Engineer | Data Center Operations & Infrastructure