Introducing Climatik: Power capping AI applications for data center sustainability

As the demand for artificial intelligence (AI) and cloud computing grows, so does the power consumption of data centers. Data centers are the backbone of the modern digital economy, but they are also some of the biggest contributors to global energy usage. With the generative AI (gen AI) explosion and the increased demand for AI workloads and power, there has also been an uptick in technology’s carbon footprint.

This presents a critical challenge: how can data centers manage their energy consumption without sacrificing the performance needed to support their AI applications? The new Climatik project, developed by Red Hat in collaboration with project contributors from Intel, Bloomberg and IBM, aims to address this challenge by offering a scalable cloud native solution that optimizes energy efficiency through power capping.

Why power capping?

Climatik introduces dynamic power capping to reduce the energy consumption of AI workloads in data centers. Based on Kubernetes, Prometheus, Kepler and Custom Resource Definitions (CRD), Climatik enables real-time monitoring and adjustment of power usage, allowing data centers to achieve a balance between performance and energy savings.

Through power capping, Kubernetes administrators can smooth the energy consumption spikes of AI workloads, leading to more sustainable operations without impacting the overall performance of these resource-heavy applications.

Technology stack

At the core of Climatik is a cloud native technology stack designed to more seamlessly integrate with existing environments. This stack includes:

• Kubernetes for managing containerized AI workloads and resources.
• Prometheus and Kepler to monitor real-time power consumption metrics.
• CRD as a method of defining power capping policies for specific workloads.
• Controller for the continuous adjustment of power caps based on metrics provided by Prometheus and Kepler.
• Webhook to enforce power capping policies dynamically.

By deploying Climatik, data centers can expect to see improvements in:

• Energy savings: Users can peak smooth power consumption and cap of overall power consumption without sacrificing performance.
• Sustainability impact: By reducing the power consumption of their workloads, enterprises can contribute lower carbon emissions.
• Performance: Enterprises can maintain the performance of AI workloads through intelligent power management.

Looking ahead

As AI and cloud workloads continue to expand, the need for energy efficient solutions has never been greater. Climatik offers a scalable, practical approach to power capping that can help data centers reduce their carbon footprint and achieve their sustainability goals. By aligning with industry best practices and leveraging cutting-edge cloud native stacks, Climatik is positioned to make a meaningful impact on data center sustainability.

In the future, Climatik aims to explore ways to scale its power capping solution to data centers and investigate integration with additional monitoring and management tools, aiming to create a flexible, scalable system that can adapt to the needs of any data center environment.

Get involved with the Climatik project via GitHub and learn more about Red Hat’s sustainability efforts on Red Hat’s Emerging Technologies blog.

To read more about Red Hat's KubeCon NA 2024 news, visit the Red Hat KubeCon newsroom.