Facing pressure to build fast & cut costs - data center owners turn to prefab & DCIM
After attending the Datacloud Europe event in Monaco last month, my #1 takeaway is that many of the technologies that have been promoted in the industry have become standard practice in the sense that they are not viewed as particularly innovative anymore but as sensible ways to deliver more efficient, more reliable, and safer data centers in a more condensed time frame.
From the various sessions and the many customer meetings at the event, a few themes emerged, including some that point to widespread acceptance of technologies including prefabricated modular data centers and data center infrastructure management (DCIM) tools.
For starters, everyone from the largest to the more niche players talked about the continued pressure to reduce cycle times, to build out capacity faster.
It’s clear that investors are urging data center owners, including colocation providers, to be more efficient with their use of capital.
They’re pushing to examine different data center designs, to make sure data centers are right-sized to quickly meet demand, without stranded capacity.
That’s why we see so much use of modular, prefabricated data center components today. Companies are realizing that you can build a data center much more quickly, and with better quality, by doing things in parallel.
While you’re getting the site ready, you can have your power, mechanical and IT components being built off-site in a factory. When the site is ready, just drop in a skid or container with all the prefabricated components.
Prefabricated data centers also make it far easier to build only the amount of data center capacity you need. For a long time, data center owners focused on power usage effectiveness (PUE) and the cost per MW.
So, they’d wind up building a 10MW data center, to keep cost per MW down, but would take 3 years or so to fill it to capacity.
That means they’re saddled with higher up-front capital costs than necessary as well as inflated operational costs, because they still have to cool and manage all that empty space.
Now folks are learning that you may well be better off building at a slightly higher cost per MW, but in smaller chunks.
The strategy works because with prefabricated modules it’s relatively easy to expand capacity as warranted; there’s no longer a need to over-build, which lowers both up front capital expenses and long-term operating costs.
Using prefabricated data center modules also helps with another issue that was brought up multiple times at Datacloud Europe: the benefit of standardization in data center builds.
Even if a module is custom designed for you, if you can use the same pieces and parts for multiple data centers, it will help optimize the supply chain, saving money and time. Also, while it may be true that each of four sites would benefit from their own unique design, the overall benefit may well be greater if you use the same design for each.
Another theme that emerged along the lines of lowering costs was the need to improve operational efficiency. This is another example of where the conversation around a technology has evolved, in this case DCIM.
In the past, the talk was all about DCIM as a technology – who’s using it, what the barriers are, and so on. Now there seems to be a level of acceptance that DCIM, especially when used in conjunction with Building Management and Power Monitoring improves data center operations and management, and the conversation has turned to how to make DCIM even more effective.
At the event, I heard lots of discussion about the role of Internet of Things (IoT) technology can play in instrumenting data center components and feeding increasing amounts of data into a DCIM platform.
There’s also talk of how artificial intelligence (AI) technology can help identify patterns in that data to lead to increased efficiency and reliability. In fact, the IoT/AI discussion was the topic of one of the event keynotes, as discussed in this previous post.
Article by Greg Jones, Schneider Electric Data Center Blog Network