Modelling out data centre energy use over the next decade
That data centres lie at the heart of digital transformation around the world is common knowledge today.
It is worth noting that various experts once predicted that data centre energy demand would continue growing and eventually consume over 20 per cent of global demand.
Fortunately, that never happened due to huge improvements in energy efficiency and the performance of new processors.
Coupled with software-based technologies such as virtualisation, containers, and microservices, and it became possible to do a lot more using the same resources to curb power consumption.
Strides in computing capability
But what will data centre energy consumption be like over the next decade? To answer this question, Schneider Electric did an internal study that attempted to model this in 2018 and forecast potential IT energy consumption. Before looking at it, it is worth highlighting some major developments within the data centre.
- Multi-core processors: While microprocessors have grown significantly more powerful over the last two decades, one development is the rise in the number of processing cores within each chip. In recent years, AMD took this to new heights with its EPYC server chips with as many as 64 cores per processor.
- Graphics Processing Units (GPUs): Once used solely for graphics processing or games, GPUs are now a vital linchpin for AI applications and specialised computing needs. As the use of AI increases around the world, expect GPUs to take on a bigger role than before in the data centre.
- HDD to SSD: There is a gradual but growing shift away from rotating hard disk drives (HDD) to solid-state drives (SSD) for storage. Some vendors have started touting the benefits of "all-flash" data centres, which are not only faster but also consumes less power.
- Quantum computing: Continuing technical advancements might soon bring general quantum computing into reality, where they can be used to tackle incredibly complex tasks that would take standard computers too long.
Need for more data centres
Ultimately, compute and storage requirements will continue to grow and drive the construction of more data centres.
Indeed, the current trend is leading to the construction of very large "hyperscale" facilities that are designed and run by specialists who operate data centres for a living. In Singapore, many of the new data centres are hyperscalers, including Facebook's mammoth US$1 billion (S$1.37 billion) facility that is currently being built.
Edge computing requirements will also rise, which are deployments located on-premises or locations nearer to where the data is generated or consumed.
This will reduce network congestion and improve latency, which can be crucial for a variety of real-time systems or AI-powered applications. In that vein, large central data centres will likely be used mainly for data storage, while processing will transition to edge data centres.
In the meantime, do check out Schneider Electric's Data Center and Edge Energy Forecast Tool to model out global energy consumption here — which demonstrates the importance of implementing best practices to minimise global energy consumption.