HEROES - Towards Eco-Responsible Hybrid HPC Marketplaces
Tuesday, May 31, 2022 9:00 AM to 6:30 PM · 9 hr. 30 min. (Europe/Berlin)
Foyer 3 + H - Ground Floor
HPC Workflows
Information
Technology is evolving quickly. While artificial intelligence and machine learning are decades old, they are just now taking off. Therefore, bridging the gap between HPC and AI/ML is more important than ever. In this context, the HEROES project will develop a software solution enabling industrial and scientific user communities to easily advance complex simulation and ML workflows to HPC and cloud infrastructures. Through HEROES, users will be able to select the best platform to achieve their goals in time, within budget and with the lowest carbon footprint – thanks to a catalogue of services provided by an HPC Marketplace.
The core of the project includes two main concepts: applying the notion of Marketplaces to HPC – where both resources providers (HPC Centres, Cloud Service Providers) and application providers (ISV, researchers, open-source projects…) can propose and advertise services; and a decision module for guided/automatic workflow placement. These embed from end-to-end the notion of eco-responsibility and will allow end-users to execute their simulations in the most energy-efficient way, while also considering the cost and time constraints. For many years, energy has been a topic only addressed from a hardware point of view, when HPC systems are designed. More recently, the rise of the electricity cost and the requirements for data centres to establish energy management as a global responsibility, made this topic not only a research topic but also and foremost a necessity for the future of the planet itself. Addressing this problem becomes even more challenging in a hybrid HPC environment.
Contributors:
The core of the project includes two main concepts: applying the notion of Marketplaces to HPC – where both resources providers (HPC Centres, Cloud Service Providers) and application providers (ISV, researchers, open-source projects…) can propose and advertise services; and a decision module for guided/automatic workflow placement. These embed from end-to-end the notion of eco-responsibility and will allow end-users to execute their simulations in the most energy-efficient way, while also considering the cost and time constraints. For many years, energy has been a topic only addressed from a hardware point of view, when HPC systems are designed. More recently, the rise of the electricity cost and the requirements for data centres to establish energy management as a global responsibility, made this topic not only a research topic but also and foremost a necessity for the future of the planet itself. Addressing this problem becomes even more challenging in a hybrid HPC environment.
Contributors:
- Benjamin Depardon (UCit)
Format
On-site
