RACA Journal May 2024 RACA_May2024_digital | Page 20

18
Business

THE COLD PLATE CONUNDRUM

( PART 2 )

Compiled by Eamonn Ryan based on a panel discussion from Vertiv ’ s website
This panel discussion hosted on the Vertiv website explored the topic ‘ The cold plate conundrum – how sustainable is it really ?’ and delved into the sustainability trends in cooling . This is part two of a two-part series .

18

Image by Benzoix on Freepik

Continued from part one . Moderated by Alex Dickens , channel management team leader : Datacentre Dynamics , the panel discussion looked at the realm of high-performance computing ( HPC ), where the question of sustainability takes centre stage , especially when it comes to cooling technologies .

The panel featured insights from three experts :
• Austin Shellnut , president and principal engineer at Strategic Thermal Labs
• Imran Latif , chief operations officer at Brookhaven National Laboratories
• Bharath Ramakrishnan , senior mechanical engineer at Microsoft
VENDOR LOCK-IN AND THE PATH TO INTEROPERABILITY The conversation delved into the issue of vendor lock-in and the importance of interoperability . Latif emphasised the need for standardisation to address potential challenges arising from proprietary solutions . Vendor lock-in , where a data centre becomes overly reliant on a specific vendor ' s technology , can limit flexibility and hinder the adoption of new , innovative solutions .
Standardisation becomes a crucial mechanism for mitigating vendor lock-in risks . By establishing common guidelines and metrics , the industry can foster an environment where different
The panel discussion provided a comprehensive overview of the evolving landscape .
RACA Journal I May 2024 vendors can provide solutions that adhere to a shared framework . This approach encourages competition , innovation and prevents undue reliance on proprietary technologies .
Ongoing dialogue and collective efforts aim to create a landscape where data centre operators can make informed decisions , optimise efficiency and navigate the challenges of evolving hardware configurations .
UNLOCKING LIQUID COOLING : A CASE STUDY IN ADAPTATION A case study was presented by Latif , shedding light on the practical exploration of liquid cooling adaptation in a data centre designed primarily for air-cooled equipment . The study aimed to assess the effects of server inlet temperature on various operational parameters , ultimately justifying the feasibility of deploying additional GPU-based clusters that required liquid cooling .
Key components of the case study :
• Benchmarking and simulation : The study involved a typical rack with 26 compute nodes , focusing on the high-energy physics benchmark — a simulation of real high-energy physics workloads . Latif highlighted the collaboration with high-energy physics collaborations worldwide , emphasising the nature of high-throughput computing and the processing of raw data from particle detectors .
• Experiment setup : The experiment maintained ambient conditions at 24 ° C , 45 % relative humidity , with an 18 ° C chilled water supply from the facility standard rack . The team ran the benchmark under three different instances , varying the conditions to assess performance and energy efficiency .
• Observations and results : The data reflected an increase in server inlet temperature to 36 ° C under the highest temperature instance , mimicking a load within the rack . Despite the elevated air temperature , the study demonstrated sustained CPU ( central processing unit ) performance with only a slight 1 – 2 % variation in the CPU score outcome . www . refrigerationandaircon . co . za