We delivered a HD computing cluster for MetaCentrum
20. 5. 2023
20. 5. 2023
The purpose of this delivery was to acquire equipment for regular renewal and strengthening of the computing and storage performance of MetaCenter clusters. The aim was therefore to acquire a computing HD cluster, which will complement and then (partially or completely) replace the computing clusters acquired by CESNET in previous years, which were due to expire in the coming months.
MetaCenter’s current compute clusters are composed of several types of compute nodes. The exact ratio between types of compute nodes changes over time based on user demands, individual cluster utilization, and technological advances in hardware accelerators. The cluster supplied by us will therefore be used for demanding scientific calculations, while, in accordance with modern methods, part of the cluster’s computing resources will be made available through a cloud environment designed not only for scientific calculations, but also for services related to these calculations. Konfigurace dodaného clusteru je přizpůsobena aktuálním potřebám uživatelské komunity – tedy primárně uživatelům z vědecko-výzkumné a vzdělávací komunity ČR (výzkumné organizace, výzkumné projekty a týmy) v širokém rozsahu aplikačních věd.
The MetaCentre was established in 1996 and since 1999 has been part of the activities of the CESNET association. It offers a virtualized environment (cloud) as the basis for a powerful computing infrastructure. CESNET, through MetaCentre, has been acting as the coordinator of the National Grid Infrastructure (NGI) since 2009 and is thus mandated to represent the Czech Republic in the European Grid Infrastructure bodies of the European Grid Initiative (EGI.eu)
Individual research teams and organisations are currently acquiring resources in the form of a computational HD cluster. By linking individually acquired computing and storage capacities into a distributed unit (Grid), it is possible to achieve a significant increase in the efficiency of investments, improve the services provided and significantly simplify the collaboration of scientists and their team at national and international level.
MetaCentrum operates and manages its own and entrusted computing resources and data storage facilities of academic centres, which are available to all researchers and students of universities in the Czech Republic. Within the MetaCentre, the central purchase of infrastructure is ensured, which is then deployed throughout the Czech Republic. Currently, the main centres of the MetaCentre are Masaryk University in Brno (CERIT-SC Centre and Faculty of Science of MU), UOCHB, University of West Bohemia in Pilsen (Faculty of Applied Sciences), University of South Bohemia (Cluster Faculty of Science), Academy of Sciences (Institute of Physics), Department of Telecommunication Technology of FEL CTU and CESNET z.s.p.o.
As part of our delivery, the computational HD cluster was divided into two locations – part of it was delivered to the Institute of Computer Engineering of Masaryk University in Brno and another part to the Institute of Molecular Genetics of the CAS.
The delivery included a total of 96 HD nodes – each representing the use of best-of-breed technologies that represent increased efficiency for MetaCentre through reduced energy consumption relative to overall computing power.
Parametr | HD výpočetní cluster pro MetaCentrum |
---|---|
Počet CPU | 192 × AMD Epyc™ 7543 |
Celková velikost operační paměti | 49 TB |
Použité CPU | AMD EPYC™ 7543, 32 jader, základní frekvence 2,8GHz (boost až 3,7Ghz), L3 cache 256MB, TDP 225W |
Celkový výkon | Výkon v pohyblivé řádové čárce dle testu SPEC2017FP rate 49 536 bodů |
Celkový počet jader | 6 144 |
Celková disková kapacita | 737 TB NVMe 4.0 U.2 |
Síťové připojení | Ethernet Converged Network Adapter X540-T2, Dual port 10GbE-T |
Max. spotřeba | ~70 kW |
Provedení | rack, 48U ve dvou vzdálených lokalitách |
AMD EPYC™7003 series processors are server processors that set a new standard for data centers. Performance you can count on to propel your modern data centre workloads. The sophisticated concept helps you to easily deploy and change workloads. These processors are suitable for virtualization, hybrid cloud or software-defined infrastructure.
These are the first server processors with x86 architecture that are built on 7nm technology. The third generation of AMD EPYC™ with the Milan label has a performance jump over the previous generation, reaching up to 25% for the most common 16-core CPU and up to 15% for multi-core CPUs (48, 64 cores).
The AMD EPYC™ 7003 generation also brings L3 cache, the biggest benefits include:
2 × L3 cache directly accessible to the core
Up to 32MB of memory per core, which can be allocated to one active core
Improved performance in applications, where more cache memory is required
Reduction of memory latency
Better performance for virtual servers
Additional benefits of third processor generation:
Zen 3 architecture that delivers increased performance per core
Enhanced security features
Support for 4, 6 and 8 memory channel configurations
Infinity FabricTN and Memory Clock Synchronized technologies that provide improved memory performance
Interested in more information about AMD EPYC™ processors? See if an AMD solution is right for your infrastructure.