5 Tips to Optimize Data Center Power Usage
An unexpected power failure can bring a data center to a halt, resulting in lost revenue and information, corrupt files, revenue disruption and equipment damage—not to mention lots of stress as you attempt to resolve the issue.
Improving data center power usage offers many benefits. It can reduce operating costs, lower energy use and extend the lifecycle of servers and other IT equipment. But it’s important for another reason, too: Data center power usage impacts reliability.
In a recent LinkedIn Live event featuring Global Director of Data Centers Chrissy Olsen, Technology Solutions Architect Henry Franc, Field Support Engineer Denis Blouin and myself, we discussed how to optimize data center power usage—and we wanted to share some of our insights with you.
Over the years, the Uptime Institute’s Data Center Industry Survey Results consistently reveal that data center downtime can most often be blamed on power outages. In 2021, 43% of data center failures occurred as a result of power outages. (Network issues, cooling failures and software/IT issues were the runners-up.)
The good news in all of this? The right processes and redundancy protocols can help prevent many data center power outages. Here are five tips we shared in our recent discussion about optimizing data center power usage.
1. Determine How Much Downtime You Can Afford
Before tackling strategies to optimize data center power usage and reduce downtime, it’s important to determine what your organization requires in terms of availability. Every minute of downtime leads to lost money and productivity. Hourly downtime costs can reach $1 million or higher for 44% of U.S. enterprises, according to the ITIC’s 12th annual 2021 Hourly Cost of Downtime Survey. How much downtime can you withstand without significant negative impacts? The answer depends on your leadership, clients, organization type and many other factors.
For example, basic email services may not have the same value as financial transactions. A few minutes of downtime that prevent email from being sent and received isn’t nearly as impactful as downtime that prevents banks and online retailers from doing business. Keep that in mind so you can determine data center availability and redundancy levels based on need instead of over- or under-designing.
2. Design for Optimal Power Usage
Data center design matters when it comes to optimizing power usage—and here’s an example. We worked with an organization that chose to build its own enterprise data center after renting space in a colocation data center facility for several years.
The new data center housed 24 racks with 3-phase 30A power in each rack. The design called for an oversized cooling system and undersized power upstream, which created an availability problem: When one rack fails, it could take the entire data center down with it.
Let that story serve as inspiration to make sure you design with availability and optimal power usage in mind.
3. Don’t Neglect Operations and Maintenance
Operations and maintenance are just as crucial as design when it comes to power optimization. Together, they create a three-legged stool: All three components are required in order for a data center to successfully stand on its own.
For example, maintenance can include regular visual inspections of systems and equipment to check for things like moisture, dust, proper illumination, noise, loose components and hot or cold temperatures that could indicate a bigger problem.
We also suggest regular failover testing—monthly, quarterly, annually and biannually—to make sure redundant back-up systems function so they can keep business running when primary equipment fails.
4. Address the Human Factor
Data center power outages can be caused by two things:
- Preventable accidents (unintentional disconnection, activation of the emergency power-off switch, etc.)
- Nefarious acts (things beyond your control, such as weather or problems with the electrical grid)
Outages caused by simple oversights or human error often happen due to the absence of processes and protocols, lack of maintenance or even people becoming complacent and assuming that shortcuts are okay.
Simple improvements like data center labeling or color coding can help. Some of our data center clients choose to use specific colors—yellow and blue, for example—to represent the A and B power feeds that support switch and servers. This makes it easy to see right away which components belong to which power feed.
Another simple mistake involves unplugging! Power cords with built-in locking devices can prevent servers from accidentally being unplugged as you reach into the rack to grab a fiber connector. Traceable patch cords can also be used to easily identify and trace network connections—and prevent accidental disconnection of the wrong patch cord.
Another pro tip: If you have technology in place to avoid unplanned downtime, be sure to use it! We’ve seen customers put significant time and effort toward maintaining uptime and availability only to miss opportunities that are right in front of them. Case in point: investing in outlet-monitored power distribution units (PDUs) but not monitoring their status!
5. Keep Power Loads Balanced
Proper load balancing—matching current draw on each phase—keeps operating costs in check and reduces downtime.
For example, we saw a data center deploy 400V 3-phase PDUs. The phases were unbalanced due to miscalculations, which blew their PDUs and caused an outage.
Load balancing is also important to consider for redundancy. If one power feed experiences an outage (A side, for example), then B side can’t become overloaded. If it does, then a breaker will trip and cause unplanned downtime. To avoid overloading, we recommend staying with 80% PDU capacity (40% per side).
Your Data Center Power Expert
No one expects you to be a data center power expert—you just need someone to turn to when you have questions. Technology and best practices change so rapidly that it can be hard to keep up.
Belden can be your resource for data center power optimization. We’re happy to answer your questions and help you achieve the availability level you need. To learn more about data center power optimization, watch our recent LinkedIn Live!
Exploring Dual-Power Feeds in Data Centers