NADIKI provides real data center information for research for the first time

NADIKI provides real data center information for research for the first time

The NADIKI project makes real, pseudonymized data from production data centers available for research for the first time—via a JupyterLab with access to infrastructure attributes, power consumption, utilization, and cooling.

The NADIKI project provides real measurement data from production systems for research for the first time. Researchers gain access to the properties of the available systems through a JupyterLab environment—data center information such as PUE, area, UPSs, and diesel generators, as well as IT system data on CPUs, GPUs, storage, and RAM—and their metrics on power consumption, utilization, cooling, network traffic, as well as read and write operations.

“With the Open Data Hub, we have already aimed to finally provide good, normalized data from data centers and IT systems for research. Now, we have succeeded in the NADIKI project." — Max Schulze

The data is pseudonymized to protect the confidentiality of operators. X-Ion, University of Paderborn, and KoloDC support the initiative as the first data centers to make their data available for research. We use the interfaces from Boavizta and Electricity Maps to convert measurement data into environmental impact indicators.

Researchers can request access to the research system via the application form.

Related publications

None

Research
NADIKI - Sustainability Indicators for AI

Within the NADIKI Project, we are integrating Environmental Impact Indicators from the Life Cycle Assessment Methodology into Kubernetes. This empowers workloads, like AI training workflows and inference, to assess their operational impact and allows detailed reporting. Using this data, we evaluate the feasibility of an environmental product declaration for an AI model.

Research
NADIKI - Sustainability Indicators for AI

Within the NADIKI Project, we are integrating Environmental Impact Indicators from the Life Cycle Assessment Methodology into Kubernetes. This empowers workloads, like AI training workflows and inference, to assess their operational impact and allows detailed reporting. Using this data, we evaluate the feasibility of an environmental product declaration for an AI model.