For decades, my engineering colleagues and I have been called upon to improve power usage in data centers of all sizes. Before we suggest any type of critical power/rack power solution, we always assess the environment with a thorough walk through, along with comprehensive testing and measuring. In these assessments we almost always see a few mistakes, regardless of the age, capacity or budget of the data center.
It’s an ongoing frustration for us to see a facility that is at best losing money, and at worst on the brink of shutdown, because of something that in some cases is easy and inexpensive to remedy. Here are the top five power-related mistakes I see in data centers just like yours.
#5 No Labeling
This is the most frequent mistake we see. In the event of a power failure, tracking down the issue is incredibly challenging if you don’t know where power is going and where it’s coming from. Without labels, you’re required to go through each PDU to see which one has a breaker that’s been shut off. That extra time puts the facility at great risk. The data center’s entire power distribution system should be labeled for troubleshooting and making changes. Creating a topographical map of the entire power train isn’t a bad idea either.
#4 Power Sharing between Racks
When you share power between racks, it’s impossible to know for sure if you’re over- or under-capacity. Sharing power equates to a higher IT loads on the server strip(s), which then increases the risk of a full outage versus a localized one. When trying to add power capacity or redundancy, three-phase UPS systems can be configured to run in parallel. For smaller loads, dual input cords may be an option, with an automatic transfer switch connecting one power source to the facility’s UPS and the second power source to a smaller, rack-mounted UPS.
#3 Daisy Chains
Daisy chaining in the data center refers to a series of power strips connecting to one another and sharing a single power input. The shortage of outlets and abundance of devices often makes daisy chaining seem like an effective solution, but the risk of blown fuses from a power surge, damaging equipment, or starting an electrical fire can be high.
#2 Not Using Metered Rack Power Strips
There’s absolutely no good reason for a data center to use consumer-grade power strips. There are lots of industrial power strip options that are only $20 more than the hardware-store ones, yet we see them used quite often. Any new equipment coming into the facility should have, at minimum, a power meter. There’s no way to do redundancy planning or provisioning with any degree of accuracy without basic data collection.
#1 Not Using Tools Already in Place
You’d be surprised at the number of times I’ve been asked to provide a monitoring solution, only to discover they’ve had the ability to monitor all along! They have purchased the best metered equipment and monitoring technology money can buy, but never use it. Most of the time this is because the back-end costs associated with hooking a system up to the network weren’t considered. Too many access points/screens/logins are another reason why monitoring systems go unused. Not only do employees get overwhelmed by all the different information access points, but they don’t know what to do with the information once it’s collected.
Is your data center guilty of one of these critical power fails? If so, the good news is that they’re all easy and inexpensive to fix. The issue is time and effort. But trust me, the benefits – short and long-term – are worth the effort!
About the Author
Rob Huttemann is the Senior Vice President of Operations for Critical Environments Group. He has more than 30 years of industry experience and familiarity with data center and supporting infrastructure management, with specific focus on power, space and storage, cooling, and overall best practices. CEG is a national provider of data center and critical environment infrastructure optimization solutions. For more information, please visit www.criticaleg.com.