A new whitepaper addresses the physical infrastructure challenges faced by data centres as part of a global shift towards artificial intelligence (AI)-driven applications, and promotes the need for regulation around AI-optimised data centre design.
Titled “The AI Disruption: Challenges and Guidance for Data Center Design,” the publication discusses how rapid developments in AI have brought about significant changes and challenges in data centre design and operation.
The whitepaper was published by Schneider Electric, which specialises in digital transformation of energy management and automation, with the company warning that ‘data centres must adapt to meet the evolving power needs of AI-driven applications effectively’.
It primarily focuses on how demand for processing power has grown exponentially with the development of AI and its growing adoption across a range of industry sectors.
“As AI becomes a dominant workload in the data centre, organisations need to start thinking intentionally about designing a full stack to solve their AI problems,” said Evan Sparks, chief product officer for artificial intelligence, Hewlett Packard Enterprise.
“We are already seeing massive demand for AI compute accelerators, but balancing this with the right level of fabric and storage and enabling this scale requires well-designed software platforms.
“To address this, enterprises should look to solutions such as specialised machine learning development and data management software that provide visibility into data usage and ensure data is safe and reliable before deploying.”
The paper highlighted that, as AI workloads are expected to increase at a compound annual growth rate (CAGR) of 26-36% by 2028, this will accordingly lead to increased power demand within existing and new data centres.
Meeting these projected energy demands will require several key considerations, especially in the areas of power, cooling, racks and software tools, according to the whitepaper.
Pankaj Sharma, executive vice president, secure power division and data centre business at Schneider Electric, added: “AI applications, especially training clusters, are highly compute-intensive and require large amounts of processing power provided by GPUs or specialised AI accelerators. This puts a significant strain on the power and cooling infrastructure of data centres.”
Sharma continued to explain that, as energy costs rise and concerns arounds these facilities’ environmental impact grow, that “data centres must focus on energy-efficient hardware, such as high-efficiency power and cooling systems and renewable power sources to help reduce operational costs and carbon footprint”.