eProsima is committed to fostering a more sustainable future. Leading the SustainML project, eProsima aims to significantly reduce the environmental impact of machine learning (ML) applications, reducing CO2 emissions and enhancing energy efficiency. By integrating the DDS communication layer into the SustainML framework, eProsima strives to optimize data sharing and enhance overall efficiency, aligning with SustainML’s ambitious environmental goals.
The SustainML project aims to create a modular and distributed framework and toolkit to enhance energy efficiency across the entire lifecycle of ML applications. It is composed of separate components that effectively communicate with each other, ensuring better scalability, reliability, and ease of maintenance compared to traditional monolithic systems.
Relying the communication layer on DDS technology, specifically eProsima Fast DDS, enables real-time communication between these components and provides the necessary communication protocols to ensure high-quality data transmission.
DDS, which stands for Data Distribution Service (DDS) technology, helps different parts of the framework share data efficiently and reliably.
Overall, the SustainML framework is designed to be flexible and efficient. It supports both local and distributed AI design scenarios while focusing on energy efficiency and sustainability.
DDS lays the foundation for the communication specification of the following key components of the SustainML framework.
🌿Front-end: This user interface allows people to interact with the framework.
🌿Back-end: Acts as the central command center, managing and collecting data from various framework parts.
🌿Task Encoder: Helps to set up ML tasks by defining parameters and optimization criteria.
🌿Hardware Constraints & Application Requirements: These components determine the hardware needs and application-specific requirements.
🌿ML Model Provider: Calculates the best ML model based on user needs and hardware constraints.
🌿Hardware Provider: Chooses the best hardware to run the ML model and offers insights for further optimization.
🌿Carbontracker: Estimates the ML model's carbon footprint on the chosen hardware.
All the SustainML participants: DFKI, Inria, IBM, University of Copenhagen, UpMem, Technische Universität Kaiserslautern and eProsima.
MORE INFORMATION ABOUT SUSTAINML:
For any questions, please get in touch with
This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101070408.