Data centres must evolve to accommodate the challenges posed by AI workloads. Salih Toyran, Director Mission Critical, chapmanbdsp, explores strategies for optimising infrastructure and implementing advanced cooling solutions while highlighting the integral role AI has in driving innovations.
As the demand for Artificial Intelligence (AI) continues to grow, data centre design and delivery must evolve to meet the specific requirements of AI workloads. Strategies to address this demand include optimising infrastructure, enhancing energy efficiency, leveraging advanced cooling solutions and ensuring scalability by fostering sustainability at heart.
Optimising Infrastructure is one of the most important aspects as AI applications require substantial computational power, necessitating the use of High-Performance Computing (HPC) systems and specialised hardware like Graphics Processing Units (GPUs) and Tensor Processing Units (TPUs).
Designing data centres with a focus on high-density computing capabilities ensures they can handle the intensive processing needs of AI workloads. This means a lot more power connected per square metre, compared to traditional data centre design. This also includes high-speed networking infrastructure to facilitate rapid data transfer between storage and computing nodes.
Increased power demand requires inevitably to think about Enhanced Energy Efficiency. The power consumption of AI workloads is significantly higher than traditional applications. Therefore, implementing energy-efficient technologies and practices is crucial. This involves using advanced power management systems, renewable energy sources and energy-efficient hardware components.
Additionally, AI can be employed to optimise power usage within the data centre itself, predicting and dynamically adjusting power distribution based on workload demands. There have been studies to create the Digital Twin of data halls to estimate and use the stranded capacity within the systems by testing the hall performance on a CFD model. We believe this can help further optimise the power demand by setting a more realistic estimation of a live data centre without requiring an upgrade of the infrastructure.
Industry should also leverage from the Advanced Cooling Solutions. High-performance AI hardware generates substantial heat, requiring innovative cooling solutions to maintain optimal operating conditions and prevent overheating. Liquid cooling and immersion cooling technologies offer superior thermal management compared to traditional air cooling. These methods can efficiently dissipate the heat from densely packed AI servers, improving overall system reliability and performance.
We need to ensure scalability is also considered as AI demand is expected to continue growing. Peace of mind solutions with modular data centre designs would allow for incremental expansion, enabling facilities to scale their capacity in response to increasing AI workloads without significant downtime or disruption.
Cloud-based solutions also offer flexible scalability, allowing organisations to leverage external resources as needed. We have been working with one of our clients to elevate the redundancy and resilience from site-level to regional-level with the help of cloud-based solutions. This helps for better utilisation of materials and resources at the site-level with marginally saved infrastructure deployment which in return helps the circular economy.
As we all know, sustainability has become very important in data centre design/operations. Utilising sustainable practices, such as green building materials, efficient waste management and the adoption of circular economy principles, can reduce the environmental impact of data centres.
We should not be focusing only ‘being green’ in operations, but rather understand our true environmental impact by calculating the embodied carbon for all steps of a data centre realisation.
In conclusion, addressing the AI demand in data centre design and delivery involves a multifaceted approach, integrating advanced technologies, and innovative practices. By focusing on optimising infrastructure, enhancing energy efficiency, leveraging advanced cooling, ensuring scalability with sustainability in mind, the industry can effectively meet the growing needs of AI while maintaining operational efficiency and sustainability.
The most encouraging idea is that AI can help us in all steps by optimising resource allocation and reducing waste, contributing to a more sustainable operation and realisation of data centre projects. Let us embrace the power of AI.
Click below to share this article
