Artificial intelligence (AI) provides an emerging set of solutions that are set to play an increasingly important role in the optimisation of data centre operations. It can enable more intelligent, real-time decisions on energy efficiency, reducing excess power consumption and carbon emissions. The technology’s ability to analyse load patterns means it can automate resource allocation to achieve previously unattainable levels of efficiency. With the ability to predict fluctuations, track system performance and identify deviations from the “normal”, AI has opened the door to the major gains delivered by preventive maintenance and enhanced cyber security.
This blog will explain how AI in data centres addresses the challenges of a data-centred society in which information volumes are growing massively, sustainability is an urgent imperative, and technological skills are in short supply.
Challenges in data centres
Data centres around the globe face many challenges as the demands on digital infrastructure grow. Density and space are a major concern and the sector’s power use, and wider sustainability impact are under intense scrutiny. With data volumes already constantly increasing, the current expansion of AI workloads requiring low latency is driving up power requirements further, placing even greater burdens on cooling systems. AI requires a high density of hardware with corresponding increases in power consumption. A single AI computer can consume 16kW to 20kW, for example.
With the growth of generative AI, surges in demand for computing power will be significant and hard to plan for. The data centre industry is also struggling with equipment failures, bottlenecks, and skills gaps, meaning there are many challenges to overcome.
The role of artificial intelligence in data centres
To address these challenges, AI-driven data centres employ applications using the technology’s algorithms to automate and transform efficiency in key areas of operation. AI works in combination with the Internet of Things (IoT), using data from connected devices and sensors in tandem with machine learning-driven applications.
The resulting automation streamlines monitoring, software maintenance and many routine but time-consuming tasks, eliminating many of the difficulties operators would otherwise face because of skills shortages.
Beyond automation, AI can use its immense powers of analysis and inference to predict the likelihood of power outages, enabling backup systems to kick in seamlessly, significantly boosting resilience for data centre customers. The intelligence data used for this analysis comes from server performance, network congestion and disk use.
AI analysis of energy and water consumption is also able to deliver reductions in both water usage effectiveness (WUE) and power usage effectiveness (PUE), enabling operators and their customers to make significant progress towards net-zero targets. With its ability to learn and act on its insights, AI can automate power optimisation dynamically according to precise requirements of servers, cooling and airflow. Critical software-enabled, AI-powered approaches are deployed at Telehouse North, where Telehouse has achieved a 461 tonne reduction in CO₂ carbon emissions as well as a 10% cooling power reduction.
Predictive maintenance with AI
All components in a data centre are worked intensively, but like any piece of equipment, they wear out. However, the process of wearing varies and it takes a huge effort to schedule replacements for every single component. If replacements follow a spreadsheet programme, however, operators are likely to be making unnecessary replacements of functioning components, inflicting costs and downtime that could be avoided.
In a data centre environment, AI solutions autonomously identify the signs of impending malfunction, enabling interventions before failure occurs unexpectedly with costly consequences. This enables remediation with minimal disruption. AI also reduces the resource requirements for monitoring and diagnosis. By constantly monitoring for critical incidents from vast amounts of data in real-time, it gives a much higher level of accuracy than any human.
Real-time anomaly detection
This ability to establish normal patterns of performance and behaviour for thousands of components and processes is one of AI’s outstanding capabilities when combined with anomaly detection.
As well as enabling predictive maintenance, this is a huge boost to security. Data centres are always under attack from hackers who constantly adapt their tactics. Detecting these new threats and exploits devised by criminals takes time and costly expertise. By recognising what is normal behaviour on a network, and by analysing masses of data in motion, AI can identify anomalous activity and trigger alerts or initiate mitigation measures. AI solutions can automatically pick up the signatures of malware and highlight gaps in security that would take data centre professionals many hours or days to detect.
Optimising workloads and resource allocation
In terms of capacity, AI algorithms enable operators to flex their requirements far faster and more accurately, scaling in line with demand to avoid performance bottlenecks. Planning for future demand is more accurate and quicker.
Software with AI capabilities is also accelerating many important tasks for workload optimisation. The technology automates the setting up and configuration of resources and networks, spins up virtual machines and allocates storage capacity, requiring only human oversight.
These solutions are capable of balancing workloads across server capacity far more efficiently than was previously possible. This reduces the risks of damaging data bottlenecks that freeze performance and inflict significant costs on customers. With data volumes exploding, AI is fast becoming critical to resource allocation, as data professionals struggle to keep pace.
As we have seen AI is transforming data centre design and how data centres operate, enabling them to meet the increasing demands of a data-driven economy where only the highest standards of performance will support competitiveness and success.
To find out more contact us at Telehouse where we are pioneering the deployment of AI in data centre operations and data centre design.