EUREKA ITEA3 AutoDC
| Reference number | |
| Coordinator | RISE Research Institutes of Sweden AB - RISE SICS North |
| Funding from Vinnova | SEK 17 113 512 |
| Project duration | November 2018 - October 2021 |
| Status | Completed |
| Venture | Eureka cluster co-funding |
Important results from the project
The project AutoDC (Autonomous Datacentres for Long Term Deployment) has developed design choices and enabling technologies for autonomous edge datacentres that can operate without human intervention despite contextual interferences like power failure and overheating. The hardware innovations are focused on the best design for autonomy. This mainly is to minimise moving parts such as fans and pumps as these require the most frequent repairs. Regarding software, the primary outcome is a reference architecture for machine learning/reinforcement learning-enabled automation.
Expected long term effects
AutoDC has led to the creation of 22 exploitable results, including eight patents on automating model management using transfer learning and feature selection. Other results include new simulation tools for optimized load balancing using reinforcement learning developed by RISE, LTH and Ericsson in collaboration, UPS-development for power grid support and immersion cooling heat rejection using no moving parts. Several collaboration that was established during the project continues, and new project ideas has already started to emerge.
Approach and implementation
The project applied a model of high work package independence, with frequent work package leader meetings to transfer results and establish cross-WP collaborations. The model has worked well, but more resources for the administrative part of the project would have been beneficial as the reporting to ITEA3 was more demanding that what foresaw in the proposal phase. It is still not clear to us how the numbers in the ITEA3-portal is updated/synchronized with national authorities (Vinnova).