Efficiency at the edge

Efficiency at the edge

Sustainability and edge computing don’t often go hand-in-hand, but as technology continues to develop, this may soon be a thing of the past. Marc Garner, vice president, secure power division, Schneider Electric UK&I, highlights some key considerations for achieving energy efficiency at the edge.

Data centres are the cornerstone of today’s digitised economy. As the demand for digital services continues to grow, more businesses are deploying distributed IT systems at the edge of the network to increase application reliability and support digital transformation strategies. However, with increased digitisation, there is an environmental trade off to be considered, mainly due to the growing amount of electrical energy required at the edge.

Various analyses suggest that today, the global installed base of IT consumes between 1,000 and 2,000TWh of electric power, equivalent to 5-10% of all power generated. Additional research estimates that by 2030, IT will consume as much as 30,000TWh.

Clearly, increasing the capacity of electrical generation in a sustainable way is an issue that is beyond the scope of the data centre industry alone, but it is beholden on such a large consumer of electricity to make every effort to increase the efficiency of its operations and to reduce its impact on the environment.

Today, Schneider Electric is committed to improving energy consumption and reducing CO2 emissions across all areas of its business. This includes reaching carbon neutrality across all of our company sites by 2025; goals to achieve net-zero operational emissions by 2030; and most importantly, to have net-zero emissions throughout our entire supply chain by 2050.

To drive efficiency at the edge, a detailed approach to tackling energy management is essential, where new and evolving technologies such as AI, liquid cooling, cloud-based software, Lithium-ion (Li-ion) and standardised infrastructure systems all play a crucial role.

Digitalisation and the edge

An inevitable consequence of increased digitisation is that data centres themselves will evolve and become more diverse. Large hyperscale facilities, often consuming hundreds of MW of power and located outside of densely populated areas will continue to host data and applications for which cost effectiveness, reliability and security are more important than speed of delivery.

Critical applications requiring ultra-low latency and increased levels of resiliency will inevitably be hosted closer to where data is generated, processed or consumed. This disruptive trend of edge computing comprises micro data centres and hyper-converged architectures, which are smaller, more specialised and located with few or no specialist technical personnel on-site. Use cases include retail branches, industrial, automotive and manufacturing facilities, and healthcare environments, all of which contribute to the Internet of Things (IoT).

The proliferation of distrbuted IT systems will in aggregate consume a great deal more power than today’s data centres. Given that each disparate edge computing site will have its own power supply, cooling solution and other infrastructure components, they will be inherently less efficient than larger data centres. In terms of energy, a hundred thousand edge facilities, each rated at 10kW, would consume 1,000MW of power for their IT load alone.

Such a significant increase in energy consumption further proves the need for edge data centres to operate as efficiently as possible, not only for economic reasons, but for the sake of the impact on the environment.

PUE and the edge

Power Usage Effectiveness (PUE) is defined as the total ratio of the overall power consumption of a computing facility to the power required by the IT equipment alone. If those 100,000 edge data centres each had a PUE rating of 1.5, the power would cost somewhere in the region of €130 million and generate the equivalent of 800k tons of CO2 annually.

Improving the PUE in each case to 1.1 would cut the cost to €100 million and the annual CO2 emissions equivalent to 600k tonnes. That would be a saving comparable to removing 50,000 petrol cars from the roads a year! In order for such efficiencies to be made, this can only be achieved through a synergy between data centre designers, operators and vendors. Power efficiency, therefore, must be at the core of sustainable data centre design.

Great strides have already been made to improve the electrical efficiency of data centres of all sizes, particularly in the area of cooling, which is the largest consumer of power apart from the IT load. Low-power computing elements, proper alignment of equipment in hot and cold aisles and purpose-built containment all help to improve cooling efficiency and reduce power consumption.

Such considerations are typically focused on air-cooled data centres, but a recent paradigm shift is the re-emergence of liquid-cooling technology. Although familiar in the days of the mainframe and in modern high-performance computing (HPC) clusters, liquid cooling is fast becoming an attractive proposition in data centres, from hyperscale to the edge.

Infrastructure considerations

Today, liquid cooling takes many forms, from hermetically sealed heat sinks, to chasis-immersed configurations. The advantage of liquid cooling, apart from its ability to permit high-density rack deployment, is that it greatly reduces, and in some cases eliminates, the need for fans, leading to immediate savings.

Standardisation, as well as the development of micro data centres, prefabricated modules and reference designs, reduce deployment times, cost and power inefficiency in facilities of all sizes. The continuous development and improvement of such repeatable designs helps to proliferate energy efficient technologies through a rapidly growing industry.

Another crucial factor in efficient and sustainable operations at the edge is next-generation Data Centre Infrastructure Management (DCIM) software, which allows remote facilities to be monitored and controlled from a centralised console as if expert personnel were on-site, a luxury that would be prohibitively expensive in many edge computing deployments.

Creating a more sustainable edge

By utilising IoT, Artificial Intelligence (AI) and data analytic capabilities built within new edge computing components, including racks, servers and Li-ion Uninterruptible Power Supplies (UPS), operators benefit from real-time visibility, allowing them to manage critical applications at the edge and minimise downtime.

Such insights can only be provided via cloud-based software, meaning today’s edge sites can be managed with similar levels of operational and sustainable efficiency that are possible in hyperscale facilities.

As edge data centres continue to be the backbone of digital transformation and a digitised economy, it becomes increasingly important that we harness their capaibilities in a sustainable and energy efficient manner. By 2020, Schneider Electric will have helped its customers to save 120 million tonnes of CO2 through its EcoStruxure architecture, such efforts are only achievable through collaborative, standardised and pre-integrated systems, those that harness the capabilities of AI, the cloud and energy efficient technology to deliver a sustainable edge.