ESCADE – Energy-Efficient Large-Scale Artificial Intelligence for Sustainable Data Centers
Improving the energy balance of data centers, AI applications, and distributed systems through optimal placement of services and applications in the edge-cloud continuum.
The ESCADE project aims to reduce the energy footprint of data centers, particularly AI applications, through the use of world-leading hardware and software technologies. To this end, a distributed AI architecture is being developed and tested using the steel production application case. This will also be the focus of Austria’s contribution.
In distributed cyber-physical systems, data collection often takes place at the edges of the system and processing takes place in central data centers with high processing capacity. However, rather than making a strict distinction between the edge and the cloud, it is becoming increasingly apparent that processing data along the entire processing and transport chain, depending on the application and resources, can be useful (edge-cloud continuum).
Performance criteria, such as computing performance and communication latency, are typically used to determine where an application should be placed in the edge cloud continuum. However, the aspect of sustainability has been neglected until now. There is a lack of suitable models and solutions that can intelligently and, as much as possible, automatically place services and applications in a constantly changing “computing continuum” while complying with necessary performance criteria and minimizing environmental and financial costs. Solving this multidimensional optimization problem is of immense importance for the ecological and economic provision and use of future highly distributed applications.
In this context, there is also a lack of measurement tools and metrics to evaluate the sustainability of distributed cyber-physical systems across subsystem boundaries. However, these measurements are essential for any modeling or optimization approaches.
In this project, Salzburg Research is therefore pursuing the following three primary goals:
- The development of a validated solution for analyzing, evaluating, and recommending optimized placement of services and applications in the edge-cloud continuum.
- The development of a tool for the automated measurement, analysis, and visualization of sustainability metrics.
- Validation of the results from 1 and 2 in the “steel production” use case of the overall consortium in a highly distributed AI architecture.
