We are pleased to announce that Cloud&Heat Technologies has been recognised by the European Commission as one of the "Key Innovators" for innovation "Energy-aware Batch Analytics / HPC Applications in Clouds" and is now listed as an innovator on the EU Innovation Radar platform. The Innovation Radar is an initiative of the European Commission that aims to identify promising innovations and innovators in EU-funded research and innovation projects.
Innovation "Energy-aware Batch Analytics / HPC Applications in Clouds"
The innovation was developed together with the Universitat Politècnica de València and Politecnico di Milano as part of the EU-funded innovation project AI-SPRINT developed. According to the EU's assessment, innovation makes a significant contribution to two of the United Nations' Sustainable Development Goals: Goal 7, which focuses on affordable and clean energy, and Goal 9, which promotes industry, innovation and infrastructure. The innovation utilises innovative tools from the AI-SPRINT project, including the Open source software Krakewhich is used to optimise energy consumption when training AI models.
The details of the innovation were published in a publication on the subject of "Greening AI: A Framework for Energy-Aware Resource Allocation of ML Training Jobs with Performance Guarantees" in June 2024 and funded by the AI-SPRINT project. A central concern of the publication is the analysis of the challenges associated with the training of Machine Learning (ML) and Artificial Intelligence (AI), especially in terms of energy consumption and CO2 emissions in cloud data centres. Given the growing environmental footprint of these technologies, an innovative approach to improving energy efficiency will be presented. As part of this approach, a comprehensive framework will be developed that integrates components from the AI-SPRINT project, including the Open source software Krake.
An illustrative example is the training of a random forest model for the classification of electrocardiograms. Here, the resources are profiled to create a precise performance model to predict the training time. In addition, the workflow is dynamically shifted to locations with more environmentally friendly energy sources. The results show that the developed framework is able to reliably predict execution time and resource estimates, making it a crucial contribution to creating a more sustainable AI ecosystem.
Technical demonstration of the innovation
In the following demonstration Patrick Thiem,Cloud&Heat project manager in the AI SPRINTS project, gave a comprehensive technical insight. This demo serves as the basis for the innovation.
You are currently viewing a placeholder content from YouTube. To access the actual content, click the button below. Please note that doing so will share data with third-party providers.
More InformationOptimisation of AI models through the use of open source software Krake
Krake is a tool for managing containerised and virtualised applications in various cloud platforms such as OpenStack and Kubernetes. It provides a standardised interface for users and plans workloads taking into account user requirements (such as hardware and costs) and the characteristics of the platform (such as energy efficiency). Krake can be used in a variety of ways, for example for managing distributed computing capacities and applications in edge clouds. The publication shows how Krake migrates a training application across different Kubernetes clusters that are operated at different locations with different energy profiles. Krake also determines the best resources based on user parameters and regularly adjusts the provisioning.
Krake was initiated by Cloud&Heat and is now part of the association ALASCA. In April 2024, it was awarded the Saxon Digital Prize 2024 in the category Open Source . Find out more: https://alasca.cloud/projects/krake/
About the AI-SPRINT project
The AI-SPRINT project (duration: January 2021 to December 2023) aims to promote innovative advances in the field of artificial intelligence (AI) for cloud and edge environments. Numerous milestones were achieved during its term. In collaboration with various consortium partners, a flexible framework was developed that not only enables the secure and efficient use of computing resources, but also takes strict data protection regulations into account. A central concern of AI-SPRINT was the development of innovative tools that support applications in areas such as Agriculture 4.0 and personalised healthcare. Joint utilisation strategies were also created to create synergies and promote the exchange of knowledge and technologies.
In October 2023, the project partners presented their successes at the final evaluation on the premises of the European Commission. Highlights include the launch of Massive Open Online Courses (MOOCs) on topics such as edge computing, docker containers and data security, which provide access to advanced AI technologies. AI-SPRINT is also present on the AI-on-Demand platform, which drives European AI research and provides tools and case studies for personalised healthcare, agriculture and wind turbines. The AI-SPRINT Studio enables developers to design AI applications more efficiently. In addition, the spin-off OneCareAI was founded, which offers solutions for stroke prevention in personalised healthcare. AI-SPRINT has thus made significant progress in AI, which has the potential to transform various areas in the long term.
Further information at www.ai-sprint-project.eu.
