The advancing digitisation and the increasing use of cloud services have triggered a real data centre boom over the past few years. New data centres are springing up like mushrooms, while existing locations are continually expanding. There is no end in sight, especially since the IoT, which is picking up speed, will lead to a real data explosion in the next few years, which will require further digital infrastructure.

One of the downsides of this generally positive development, which makes new products and services possible, is the growing energy demand in data centres, which not only causes high costs, but also makes it difficult to achieve climate and sustainability objectives. According to the Borderstep Institute, German data centres consumed 16 billion kilowatt hours last year - a whole billion more than in 2019. Data centres have become significantly more efficient - as shown, for example, by ever-lower PUE (Power Usage Effectiveness) values.

The PUE value sets the total energy consumption of a data centre in relation to the energy requirement of the data centre IT, i.e., it shows how efficient the basic infrastructure is, including cooling systems and pumps, UPSs and batteries. Numerous improvements such as optimised cooling concepts, extensive use of waste heat, more robust IT components that no longer require extremely low room temperatures, and well-coordinated current transformers have ensured that new data centres today usually achieve a PUE value of 1.5 or lower. More than two-thirds of the energy is actually consumed by the IT systems.

With ten billion kilowatt hours, these accounted for most of the electricity consumption of data centres in 2020, but the hardware is now extremely energy efficient. According to the Borderstep Institute, their energy consumption has increased by 75 percent since 2010, while their performance has grown eightfold. IT manufacturers' many technical developments have contributed to this, but also stricter legal provisions such as the most recent Regulation (EU) 2019/424, which, for example, stipulates a minimum efficiency level for power supply units in server and storage systems.

MORE PERFORMANCE WITH THE SAME POWER BUDGET
Close interaction of the servers, storage systems and network components in the data centre makes their energy consumption interdependent. Increasing amounts of data not only lead to higher power consumption on the storage side, but also on the server and network side, because the data usually has to be transmitted and processed. Only in the case of archived data could the storage systems be viewed independently in terms of energy conservation, but validations that use other IT components are also carried out on a regular basis.

Data centre operators therefore plan a certain power budget for a rack, which is allocated between the individual consumers. For applications with very high memory requirements, however, it is important to take a closer look at storage media efficiency, because every watt that does not have to be used for their operation is available for other systems in the rack. By using energy-efficient storage, the rack can achieve a higher level of performance with the same power budget.

SSD PERFORMANCE COMPENSATES FOR POWER CONSUMPTION
In theory at least, SSDs require less energy than hard drives because they do not have any mechanical components. However, because SSDs have to perform certain operations such as managing memory cells and updating memory states in order to remain ready for use, when idle they consume just as much power as hard drives. In operation mode, their power consumption is significantly higher than that of HDDs. For example, while a current fourth generation PCIe SSD from KIOXIA has an active power consumption of up to 21 watts, an Enterprise HDD with 7,200 revolutions per minute requires less than half that.

However, this consideration neglects the performance of the two storage media types. With its 8 to 12 watts, the hard drive only provides for a few hundred IOPS, whereas the SSD achieves up to 1.4 million IOPS. This means that SSDs are much more energy efficient than hard drives in active use. They provide data much faster and therefore run for a much shorter period of time, with maximum power consumption at full load for a given workload.

For example, a KIOXIA CM6 transfers data at 6,900 MB per second and only needs 72 seconds for a 500 GB file, which corresponds to work of 0.4-watt hours with a maximum power consumption of 21 watts - the actual power consumption is somewhat lower for sequential read operations. A hard drive, on the other hand, needs around 28 minutes to transfer the 500 GB and converts 4-watt hours with a power consumption of 9 watts - the SSD is ten times more energy efficient. However, their real strength lies in random access. Compared to the approximately 250 IOPS of a typical Enterprise HDD, the 1.4 million IOPS of the KIOXIA SSD ensures 2,400 times better energy efficiency.

BETTER COOLING THROUGH NEW FORM FACTORS
As a result of their high performance, SSDs also have advantages in terms of cooling. It is true that an SSD of the latest PCIe generation has a higher absolute cooling requirement in active operation at full load than an HDD in the same period of time. However, the hard drive would take significantly longer to transfer a file or a certain amount of IOPS and would therefore have to be cooled over a longer period of time, so that it ultimately requires higher cooling capacity.

In addition, state-of-the-art Enterprise SSDs are increasingly relying on Enterprise & Data Centre SSD Form Factors (EDCSFFs), in which the flash memory modules ensure better access of the cooling airflow than the classic 2.5-inch form factor. This allows the waste heat to be better dissipated, which provides for more efficient cooling of the increasingly powerful storage media and is another small step on the way to making data centres climate neutral by 2030, as required by the EU.

More info: www.kioxia.com