The future of data centre operations will be shaped by a greater reliance on on-premises artificial intelligence (AI), a heightened focus on sustainable practices, and the implementation of AIOps, according to Perry Sui, Area Vice President, ASEAN and Taiwan, Juniper Networks. He forecasts these themes as crucial data centre networking trends that are anticipated to materialise by 2024.

Speaking about the growing shift towards AI, Sui projects that on-prem AI data centres will likely be favoured for enhanced control, security, and cost reductions, reflecting lessons learnt from past experiences with the rush to public cloud adoption.

"To cloud or not to cloud: AI version. We've seen this movie before. Five or ten years ago, enterprises started rushing to the public cloud, enticed by utopian promises of greater flexibility and lower costs," Sui explains. "However, most companies eventually realised that public cloud is not as simple and cheap as so many had thought. Cloud regret is the new story as we hear countless anecdotes of companies repatriating workloads back to private, on-prem data centres."

Given these circumstances, Sui proposes that businesses are becoming more cautious and thoughtful about decisions regarding new AI data centre infrastructure, predicting that many companies will choose to develop their on-prem AI data centres for better control, increased security, and, yes, reduced costs than the alternatives provided by public cloud.

In addition to the shift in AI focus, the geographic location of data centres is predicted to rise in importance due to a rapid escalation in power consumption. Sui explains, "Infrastructure vendors will continue to design and build more efficient gear. But driven by the rigorous demands of new AI model training, data centre racks will continue to consume more power from 10kW to over 100kW for racks in some cases."

Sui suggests that to meet these power requirements, providers will need to consider aspects such as access to renewable energy types, cooler climates, and experimental cooling methods for optimum sustainability.

Lastly, Sui expects that AIOps, or Artificial Intelligence for IT Operations, will start to become more pervasive in the data centre environment, particularly in relation to predictive maintenance and troubleshooting. If any issues do arise, AIOps can streamline the troubleshooting steps normally executed by a network operator, thereby reducing the mean-time-to-repair and mean-time-to-innocence.

To support this, AI-based large language models (LLMs) will likely be integrated into virtually every interface, enabling operators to obtain rapid insights into areas such as the current network state or recommended upgrades. As Sui explains, "Tools will look for patterns that typically foreshadow problems and proactively notify IT with the changes needed to prevent performance degradation or outage." This approach not only speeds up the process but also significantly improves the overall operation and management of data networks.