Enhancing energy efficiency in a data-intensive IT world

As we push for smart nations, are modern energy sources able to keep powering current data centers? 

Innovation in clean energy and automation will enable data centers to continue running with efficiency and performance. There are companies who have demonstrated that this is possible. For instance, web giants such as Facebook, Baidu, and Google have paved the way in this regard through power purchase agreements of solar and wind energy. 

Together with innovative, connected technologies around IT, power, and cooling infrastructure, we believe energy can be used in an efficient and sustainable way. While currently these practices are being implemented with large scale data centers, they will eventually be adopted by smaller enterprise centers.  We see the adoption of renewable energy and innovative technologies as key to further improve the energy landscape. 

Are we developing alternative energy or modern energy generation solutions quickly enough to keep pace with demand? As we drive demand up, can we keep on producing enough energy to meet the demand?  

There are opportunities to generate more energy to meet increasing demands through large deployments of solar and wind farms throughout the world, as well as bringing on new power plants. However, it is not just about creating more energy sources, it is about managing the demand side of the equation as well. In the United States, there are incentive programs that encourage data centers to use their battery energy storage as a means of curtailment.  

In the future, we will need more storage capacity to meet the growing demand for energy, but there are opposing forces to the creation of this additional capacity. On the one hand, the demand for data is continuously increasing, and on the other hand, there is a new generation of IT solutions that optimise our energy efficiencies, allowing us to maximise our output for every watt.

We are seeing this increasing demand in power and cooling architectures as well. For example, there are chillers today that can drastically increase economiser hours by supplying 20 Celsius of chilled water. Older chillers would not be able to operate at such a temperature. 

What about cooling? Are we addressing it correctly in Singapore? Are the BCA standards enough to meet green or efficient energy standards? Can Singapore truly push itself to the forefront of being a data center hub?       

Although Singapore has a challenging climate due to strong sun exposure and humidity, there are energy efficient solutions that can increase cooling efficiencies while maintaining environmental sustainability. For example, chillers can be used to increase economiser hours. Indirect air economizers, if water is available, can also be used for evaporative cooling. Admittedly water usage is becoming a bigger issue. Another solution that may hold promise is immersion liquid cooling, which could achieve 100% free-cooling even in Singapore’s tropical climate.

Is PUE still relevant as a measurement of data center efficiency? Are there other things we should be looking out for? What should we be aiming to achieve in Singapore? 

Yes, Power Usage Effectiveness (PUE) remains a key metric, but Water Usage Effectiveness (WUE) is as important. In the early days, there was a task force trying to come with metrics for IT work produced – per Watt of power consumed including kilobyte of stores/watt, number of web pages served/watt. This was the best measure of efficiency because it forced the IT department to be more efficient. Unfortunately, it is very difficult to standardise metrics around this idea. For instance, eBay has implemented this on their own by developing a dashboard of metrics that works for their organisation. 

Another crucial aspect is achieving carbon reduction goals. Countries around the world, including Singapore, have taken steps to reduce greenhouse gases in line with the Paris Agreement to lower their carbon footprint. We are at the forefront of sustainability at Schneider Electric and our solutions, such as air economisers and chillers, enable data centers to meet these goals.

Over the past few years, as the Cloud took hold and shape, was there a drop in data center demand/capacity? Can’t we simply fill up the space as demand increases? 

I think you’re referring to the dot-com crash in 2000. If so, this was a time of great expansion, where many colocation data centers were empty and later acquired for pennies on the dollar. But all of that slack has been taken up over the last decade. We are seeing the growth again and this time the colocations are being filled. 

As we roll out IoT and digital transformation, are data centers able to meet network and traffic demands? Are current DCIM solutions adequate?  

I posed this question to a Gartner research analyst a few years back and he had no doubt that there is enough bandwidth available. As the data demand increases, they just lay more fiber. To manage the growth of data centers at the current pace requires a cloud-based DCIM. At Schneider Electric, we have EcoStruxure™ IT – a cloud enabled DCIM software that can monitor data centers and IT locations.

The appeal of a cloud based DCIM is tied to a central tenet of best practices for data center operations – monitoring. Without monitoring, there are no benchmarks for improvement and therefore no possibility of optimising data center capacity and resources. The adoption of cloud-based DCIM allows companies to leverage the low costs and benefit from the data driven approach.

How much legacy infrastructure is still in data centers in the region? Are they going to be able to modernise fast enough?  

It is key for companies to modernise legacy infrastructure to support the latest IT hardware and increasing bandwidth requirements. Singapore, where data center capacity is one of the highest in the region at 370 megawatts, is the most robust Asian market in terms of business operations, based on a study by Cushman and Wakefield.

With the growing amount of data being generated and a shift towards outsourced data centers, companies with legacy infrastructure in their own data rooms are able to catch up and consider alternative options, as well as redevelop their data centers to maximise capabilities.

What are enterprises still overlooking or neglecting in data centers?

One thing I often notice are data centers that are designed to a specific rack density (i.e. 7kW/rack) but deployed at an average below that. Deploying a lower average rack density results in stranding power and cooling capacity. While it is possible to design for the capability of 7kW/rack, businesses should build a larger floor space to accommodate more racks in case the average falls below 7kW. Another way is to mount using concrete to lower costs. 


Victor Avelar, Director and Senior Research Analyst, Data Center Science Center, Schneider Electric