Project Motivation
The increasing demand for computing, particularly driven by HPC and AI, is fueling the expansion of supercomputers in both number and scale. This growth comes with a significant rise in energy consumption. As a result, supercomputer centers across Europe are facing significant sustainability challenges.
Despite advances in energy efficiency, computing demands are escalating at a faster rate, making sustainability a pressing concern. Furthermore, societal concerns about the environmental impact and high energy costs underscore the need for improved efficiency.
In this context, enhancing energy efficiency is a critical enabler for Europe’s broader digital and green transition.
Project Objectives:
The project’s ultimate goal is to develop a unified European software suite that can be seamlessly adopted across diverse and heterogeneous supercomputing architectures. To support widespread take-up, the project builds on the competency of well-established research groups, HPC/AI centers and companies, and extends widely used open-source codes.
By doing so, SEANERGYS will help reduce energy waste and operating costs and at the same time maximize the scientific and industrial benefits for Europe’s investments in HPC and AI infrastructures.
By developing an innovative chain of holistic supercomputer and facility monitoring, advanced, AI-based data analytics, and dynamic system management and scheduling the EuroHPC project SEANERGYS is helping to reinforce the European commitment to sustainability while making progress on long-term benefits across scientific research, industrial innovation, and societal advancement.
SEANERGYS addresses these challenges by providing a comprehensive software suite capable of optimizing the energy-aware operation of EuroHPC supercomputers in real time, reducing energy consumption and balancing this with the needs of the supercomputer users. The four core objectives are
- reduce the energy consumed by real-world HPC and AI workload sets
- ·improve resource utilization
- enhance overall system throughput
- optimizing response times.
Project Approach:
SEANERGYS will achieve its objectives through the combination of three interconnected pillars:
· A Comprehensive Monitoring Infrastructure (CMI): collecting a detailed system and facility data to provide a holistic view.
· An Advanced Artificial Intelligence-based Analytics System (AIDAS): analyzing monitoring data (including performance and energy metrics) and using AI/ML methods to produce insights on possible optimizations and predictions of workload and system behavior.
· A Dynamic Scheduling and Resource Management System (DSRM): utilizing monitoring data and AIDAS insights to optimize workload execution and system operation in real-time.
The SEANERGYS software suite will also harness the adaptability of dynamic workloads to minimize energy consumption and maximize throughput. By integrating the three pillars via a coompn data plane, SEANERGYS will enable unprecedented levels of coordination and optimization for HPC/AI system operation, surpassing the current state of the art and paving the way for efficient support for future workloads.
Project Background:
SEANERGYS has been selected as a result of the call HORIZON-EUROHPC-JU-2023-ENERGY-04 and will last for 48 months.
The project is coordinated by Forschungszentrum Jülich GmbH and brings together 16 partners from across Europe to develop and validate the fully integrated, production-ready SEANERGYS software solution for improving the energy efficiency of operating HPC and AI supercomputers..
SEANERGYS has a budget of €33 million and funded by the EuroHPC Joint Undertaking, which receives funding from the Horizon Europe programme and national agencies of .
SEANERGYS builds upon results of several previous projects funded by the EuroHPC JU, including the DEEP and SEA projects and REGALE.
The full press release from the EuropHPC Joint Undertaking can be found here: New EuroHPC Project SEANERGYS Advances Energy-Efficient Supercomputing – EuroHPC JU
SOURCE: SEANERGYS Project Launched: European Software Advances Energy-Efficient Supercomputing