Gauss Centre for Supercomputing hat dies direkt geteilt
Energy consumption is the biggest challenge for (scientific) #data centres and in #supercomputing: seven data centres, including the LRZ, and researchers from various universities have compiled their experiences and recommended strategies. The most important points: 👆 AI and growing #HPC systems are increasing energy consumption and therefore costs: strategies to reduce power requirements are needed. 👆 These approaches must not be limited to the hardware and #software, but must affect the entire data centre. 👆 Direct liquid cooling is becoming standard in supercomputing, and the waste heat can be passed on to heat-intensive institutions. 👆 #Hardware is becoming increasingly heterogeneous as components work more efficiently through specialisation. 👆 Heterogeneous systems require technologies and tools to flexibly control the use of hardware and the order situation: Node sharing and dynamic planning mechanisms are the keywords. 👆 Comprehensive #monitoring for the #data centre creates the basis for optimising the locations for computer systems. 👆 The local HPC sites are currently investing money and time in #user support: #application codes and software are being trimmed for higher performance, and optimised libraries, programming models and tools are also being created to increase #energy efficiency as well. To the paper: https://lnkd.in/dc9Va6P6 #IT4Science, #HPC4Germany Gauss Centre for Supercomputing, EuroHPC Joint Undertaking (EuroHPC JU), ETP4HPC European Technology Platform for HPC, PRACE (Partnership for Advanced Computing in Europe), HPC-AI Leadership Organization (HALO) Central European Chapter of Women in HPC, WOMEN IN TECH ® Global