Meluxina
LuxProvide’s supercomputer MeluXina was inaugurated in June 2021. MeluXina was built to serve a large variety of complex, data-driven workloads. Its design is forward-looking, responding to the convergence of simulation, modeling, data analytics and artificial intelligence.
MeluXina is based on the EVIDEN BullSequana XH2000 platform with a computing capacity of 18 PetaFlops per second, and counts on 20 PetaBytes of DDN storage. MeluXina is fully scalable due to its open and modular architecture, and delivers high sustained performance through its GPU AI accelerators.
Meluxina legend
The love story between the mermaid Melusina and Count Siegfried of Luxembourg is one of the country’s founding legends. LuxProvide’s supercomputer was named MeluXina after the magical creature, as it is water-cooled and also represents a founding brick in the Luxembourgish digital ecosystem.
WHAT IS A SUPERCOMPUTER?
A supercomputer is a powerful and advanced compute platform that is designed to perform complex calculations and process vast amounts of data at extremely high speeds. It represents the pinnacle of computational power and is used for solving large and complex problems in various domains.
Supercomputers are typically composed of thousands or even millions of processing units (CPUs, GPUs), interconnected by a high-speed network. They leverage parallel processing techniques, where multiple processors work together on different parts of a problem simultaneously, solving large and complex problems in shorter timeframes. This enables organizations to tackle computationally intensive tasks that would be infeasible or time-consuming for conventional computing environments.
Supercomputers are known for their exceptional processing speed, measured in floating-point operations per second (FLOPS) or more recently in terms of AI-specific operations per second (AIOPS) for machine learning workloads. The world’s fastest supercomputers can perform calculations at speeds in the range of petaflops (quadrillions of calculations per second) to exaflops (quintillions of calculations per second).
LUXEMBOURG’s NATIONAL
SUPERCOMPUTER MELUXINA
has been build to server a large variety of complex, data-driven computational workloads. Its design is forward-looking, responding to the convergence of simulation, modeling, data analytics and artificial intelligence. This enables predictive analytics simulations.
CAPABILITY
90.000
HPC CPU Cores
5.120
Cloud vCPUs
800
GPU-AI accelerators
+450
TB RAM
20
Petabytes high
performance storage
300+
Tailored software
packages
PERFORMANCE
COMPUTE
18 FP64
PetaFlops
500 AI
PetaFlops
DATA
500+
GB/s Scratch storage
190+
GB/s Project storage
INTERCONNECT
200
Gb/s on Cluster nodes
400
Gb/s on Accelerator and
large memory nodes
When it was launched in June 2021, the MeluXina accelerator module has been ranked 36th in the Top500 World ranking of supercomputers, and 4th greenest in the world and greenest in the EU.
TOP500
36th
GREEN500
4th(world)
1st(EU)
FLOPS PEAK PERFORMANCE
If all the humans on the planet were to make a calculation in one second, it would be necessary to
multiply the population of the Earth by more than two million to reach MeluXina’s capacity
MELUXINA ARCHITECTURE
SOFTWARE STACK AND CONTAINERIZED WORKLOAD
MeluXina has a rich software environment, enabling scientists and engineers to be rapidly operational and productive. It comprises hundreds of ready-to-use software tools for running scalable HPC, HPDA and AI workloads and for application development.
MeluXina’s software environment is built around several pillars:
Compilers, programming languages and tools for performance engineering
Libraries for accelerated and scalable computing, and data analytics
Frameworks,including machine learning and AI, runtime and platform tools
Domain-specific applications for use cases in precision medicine, computer-aided engineering, materials design and many other advanced fields
The software environment is optimized for the MeluXina supercomputing platform and updated on a regular basis.
MeluXina can also run containerized applications and application stacks, with support for parallel and accelerated workflows. Docker and OCI containers are supported by the HPC-ready runtime, as well as running encrypted and trusted containers.
RICH AND OPTIMIZED SOFTWARE STACK
Packages used for scalable AI
- Frameworks and dependencies already configures and available
- GPU-accelerated and optimized versions
+ 300 tools
Available on MeluXina
MELUXINA USER SOFTWARE ENVIRONMENT
Compilers,
Languages &
Performance Eng.
Parallelization tools,
MPI suites &
acceleration libraries
Numerical &
data libraries
Frameworks, runtime
& platform tools
End user
applications
MELUXINA AND CONTAINERIZED WORKLOADS
Singularity-CE
Container system on MeluXina
- Docker & OCI compatible
- Non-privileged mode for
improved security - Support for GPU accelerated
applications - Support for creating and
running encrypted containers - Support for trusted containers:
PGP signed & verified
Bring-your-own-spoftware stack
- Enable users to easily control
complete stack – REPRODUCIBILITY - Users create tooling and pipeline on
their infrastructure – RUN PIPELINE
ON MELUXINA
USING MELUXINA
MELUXINA, THE CLOUD-ENABLED SUPERCOMPUTER
We are engineering the MeluXina Cloud to provide gateways to our supercomputing platform, supporting your most intensive computational and data workflows.
The MeluXina Cloud is meant to host:
- interactive environments for building and running AI models
- container registries that deliver encapsulated, trusted applications to supercomputing container engines
- portals and APIs that will enable you to deploy applications tailored to your needs
- object-oriented interfaces to our data pools
A Kubernetes platform running on top of the MeluXina Cloud is in the works. It will link the supercomputing platform to flexible, cloud-native services, and empower scalable AI solutions.
MELUXINA, THE CLOUD-ENABLED SUPERCOMPUTER
We are engineering the MeluXina Cloud to provide gateways to our supercomputing platform, supporting your most intensive computational and data workflows.
The MeluXina Cloud hosts:
- interactive environments for building and running AI models
- container registries that deliver encapsulated, trusted applications to supercomputing container engines
- portals and APIs that will enable you to deploy applications tailored to your needs
- object-oriented interfaces to our data pools
A Kubernetes platform running on top of the MeluXina Cloud is in the works. It will link the supercomputing platform to flexible, cloud-native services, and empower scalable AI solutions.
System Documentation
Documentation
gaining access, getting started,
technical documentation)
Check the system status
