Big data’s dirty little secret: a burning need for data center efficiency

This audio was created using Microsoft Azure Speech Services

The technology trends that promise to improve the level of information we have at our disposal to make decisions—big data and the Internet of Things (IoT)—have an under-appreciated downside to them: added energy consumption in data centers.

For big data to be of use, it needs to be processed and analyzed, and all that “crunching” of data is going to take place in data centers. The resulting challenge—which was noted in a recent presentation in Singapore—is that unless data centers become more energy efficient and flexible, the big data trend is going to carry with it unnecessary energy burn, some of it from electricity generated from fossil fuels. Big data has great potential, but we need to avoid any corresponding increase in energy waste.

According to estimates from Gartner, the global information and communications technology (ICT) industry already accounts for about 2 percent of global carbon dioxide (CO2) emissions. Looking just at data centers, Gartner estimates that energy accounts for 12 percent of all data center costs, and is the fastest rising cost.

If all the energy used by data centers actually went into computation, there would be very high efficiency in the energy spent, but that is not the case. The ratio which indicates how much energy gets applied to computing tasks and information technology (IT) assets is measured by the power usage effectiveness (PUE) ratio. An ideal PUE—one where all the energy was applied to computing tasks—would be 1.0. Studies on the average PUE of data centers have varied, but we know that despite well publicized PUEs at a few data centers of barely above 1.0, the average is likely closer to 1.8, meaning that there is considerable room for improvement in data center energy efficiency.

As governments in places such as Singapore look to accelerate the availability and use of big data (Singapore’s “Smart Nation” plan proposes “above ground” boxes to relay data from public infrastructure such as traffic lights), the need for better data center efficiency will increase. Virtualization has emerged as a way to squeeze more utilization from server hardware, but aside from virtualization, one of the best ways to improve data center efficiency is to reduce the energy consumed by power, cooling, and lighting infrastructure.

As a global specialist in energy management with solutions for data center physical infrastructure (DCPI), Schneider Electric can help make data centers much more energy efficient. Part of the answer will be in planning and designing very large, highly efficient data centers that push PUE down close to 1.0. But there are also existing data centres, which through assessment and a lifecycle approach to data center services, can become much more efficient and can help meet part of the demands brought on by big data.

There are many specific tactics and upgrades that can help, including lower data centre operating temperatures, use of hot-aisle/cold aisle configurations with better air containment, equipment with “economizer” modes of operations, and use of data center infrastructure management (DCIM) software. Assessments and baselining can help pinpoint what’s needed most.

There also is the need to think about methods of continuous improvement for making data centres greener, and new approaches to making data centres more energy efficient by using data science and predictive analytics. Schneider Electric sees data science services as way to make data centres nimble enough for the massive computing demands that will be brought about by big data. Data science can also be tapped as a way to help data centres improve under Singapore’s SS 564 green data centre program, which gives incentives to operators who adopt an effective framework for continuous improvement.

We know that new approaches and services are needed to ensure that the market has data centres which are large enough, flexible enough, scalable, and energy efficient enough to cope with the coming demands of big data. Relying on the same techniques as in the past and hitting merely average levels of energy efficiency is not an attractive option when you consider the rapid growth in big data. While it’s hard to pin point the size of the coming big data explosion, a recent IDC forecast predicts that big data technology and services will grow at a 27 percent compound annual growth rate through 2017, or about six times faster than the overall ICT market. If we don’t do more to address data center energy efficiency, that sort of growth curve for big data is going to mean a sustainability hit for the environment.

Note: Thanks for to my Schneider Electric colleagues Brian Parkinson and Joycelyn Longue for helping to form and review the content for this post.


Tags: , , , , , , , , ,