Optimizing Space and Power in a Data Center

Posted on May 24, 2021 by Gento

Constant change in how consumers and organizations use IT impacts how businesses conduct their activities and operations. At the same time, IT technologies are evolving at a fast pace. For those responsible for providing the compute firepower, these two factors mean one thing: IT has to operate at the speed of business – any slower and the business suffers. 

For IT to match the pace of business, its core team has to stay on top of the game in order to be able to adapt to rapid change on both the business and technology fronts. Staying current will enable it to continually optimize the IT core – the data center. And almost without exception, such optimization has to be done within the physical confines of the data center.  

Every data center is divided into grey floor space and white floor space. The former is where back-end infrastructure (switchgear, UPSs, chillers, etc.) is located; the latter houses IT equipment such as servers, racks, storage, and power distribution system. The grey to white ratio in a typical data center is 30:70. Racks take up about 45% of white floor space – hot and cold aisles use up the remainder – meaning less than 30% of total data center floor space is actually productive.  

Notwithstanding the Covid-19 pandemic, floor space still comes at a premium so even adding more to a data center is an expensive option, let alone building a new and bigger one. Additionally, businesses, their customers and regulators are becoming increasingly climate-conscious and data centers do consume large amounts of energy. Faced with these constraints, businesses wanting to optimize their data centers are choosing to do so by going in one direction – up. 

Given that only less than a third of data center floor space supports the business and contributes to its growth, it is hardly surprising that businesses are going vertical. This could take the form of simply increasing rack height (subject to load bearing capacity) and repositioning gear like patch panels and network switches over the racks. Also, more often than not, there is unutilized space available in every rack. This underutilization is not uniform across racks so it’s a matter of looking into each and seeing what can be rearranged and added.  

As with rack space, power within a rack is usually underutilized and can be optimized. Here, the use of a busbar or an intelligent power distribution unit can give a true picture of how much power is consumed within a rack in real time and within time periods. Such information can be used to identify non-working processing assets, identify those that draw disproportionately more power for little computational output, help energy efficiency planning, and assist in capacity planning for power protection, distribution and cooling infrastructure equipment.  

As for data centers with a single-phase power supply that are near the ceiling of their power capacity or are running out of rack space, one very sensible solution is to switch to a three-phase supply that can enable more processing assets to be placed with each rack.  

All in, investing in vertical growth in the rack is far more beneficial in the long term than doing a retrofit.