Intelligent Data Centres Issue 25 | Page 20

DATA CENTRE PREDICTIONS
amounts of data , or ML training sets . For example , millions of faces could be scanned to provide an algorithm with a comprehensive understanding of the nuances of human expression . Once created , these completed algorithms could be tasked with reacting to massive amounts of real-time information such as facial tics , furrowed brows and pupil dilation .
AI / ML data is typically housed in vast data lakes . Specialised servers equipped with accelerators , GPUs for example , are ideally suited to processing AI / ML tasks . Data centre networks are ramping up bandwidth to feed these systems with very large data pipes , enabling the costeffective development of AI tools .
IoT deployments will ramp
IoT applications are rapidly proliferating as companies seek to better manage facilities and occupants . Newer connectivity protocols like LTE-M and Zigbee are enabling wireless sensors for temperature , water use , room occupancy , HVAC control and other applications , while Power over Ethernet ( PoE ) is enabling everything from Wi-Fi access points to surveillance cameras .
IoT provides critical data that drives the optimisation of manufacturing for example , feeding a trend to apply AI to process controls . In cases where the communication is between machines , data communication systems must provide very low delay or latency . Latency is a primary reason that new smaller distributed systems or Edge data centres are deployed in close proximity to their supported systems . This trend is accelerating the deployment of distributed network facilities to support a large number of Edge data centre applications .
As IoT applications continue to multiply , the amount of data that will be generated is expected to grow exponentially . Processing this data locally , close to the edge , is perhaps the most effective way of dealing with IoT data . Gartner has predicted that approximately 65 % of all servers will be deployed in Edge data centres by 2025 .
The drive to single-mode fibre
Remote workers and shoppers demand immediate response times and this will drive widespread adoption of singlemode fibre . Single-mode fibre has been
around for years , but as data centres ramp up adoption of 400G Ethernet in 2021 , we will see deployments accelerate . Adoption was somewhat slowed in 2020 due to the difficulty of obtaining components from China , but this is expected to change this year .
Data centre capacity must continue to grow , however , there must also be a continuous improvement in data centre efficiency . This is precisely why fibre networks are shifting the bandwidth of network optics up – creating a need for more efficient network switching elements and driving the use of ‘ fibre to the server ’ as previous generations of copper cabling reach speed and distance limitations . The IEEE 802.3db task force is targeting 100 , 200 and 400Gbps speeds for short reach server connections which will aid the development of lower cost VCSEL based optics .
Accommodating remote workers and customers , making facilities safer and more efficient , and driving higher performance will be the hallmarks of data centre trends for 2021 . Companies that pursue these initiatives will be at the forefront of Digital Transformation as the industry ’ s evolution continues . ◊
20 www . intelligentdatacentres . com