High Performance Computing
The E4 solution for high performance computing
A highly flexible HPC structure: you can add new nodes or interface with the cloud in a single infrastructure layout, without having to modify the systems architecture
A look to the future
Medooza is the HPC infrastructure engineered based on our twenty years of experience in the academic and industrial fields and designed with the future in mind
“Jellyfish are free-swimming marine animals with umbrella-shaped domes and tentacles as extremities.
The hood can pulsate to provide highly efficient propulsion and locomotion.
The tentacles are armed with stinging cells and can be used to capture prey and defend itself from predators“
Wikipedia – Jellyfish
Medooza offers HPC clusters the intuitive, flexible and easy to maintain approach of traditional virtualization and cloud infrastructure. You can now adopt an HPC solution without the need to fill in missing specialized skills.
Both for hardware and software. Supports the widest possible range of CPUs, accelerators, high performance networks and the most widely used operating systems
The heterogeneous development environments becoming flexible and reconfigurable. Their management is simplified through the micro-services infrastructure and native support for HPC containers
A single point of contact to support your cluster and your storage, from hardware to applications. Focuses your team’s resources on simulations and not on systems management
Maintaining and updating an HPC system, adding new nodes and interfacing with the cloud has never been so easy! And it does not require any changes to the systems architecture
Discover the advantages
Choose the size for you
The logical architecture of the E4 HPC Cluster is based on a single infrastructure layout in which the software functionalities to manage and optimize the cluster itself are deployed in natively containerized format. The initial resource sizing phase for the resources required for workload processing will identify the most suitable size from among those available: Small, Medium, Large.
Furthermore, all the components are individually tested to provide maximum performance homogeneity. After integration, complete systems are tested for a minimum period of 120 hours to ensure a perfectly functioning system, eliminating DoA (Dead on Arrival) situations.
Ease of use
Manage all aspects related to identity management, authorizations and everything related to the security of the cluster in a simple and centralized manner.
HPC container support simplifies application version control, dependency management and workload execution, providing a seamless transition to Docker & Kubernetes.
Performs the deployment of dynamic environments, capable of adapting to the most diverse workloads using personalized images for the main Linux distributions. We know that each project has different requirements, now you can finally forget about these complexities using images which are ready for deployment and easily personalizable.
Universal software stack
Supports ARM, PowerPC, x86_64 ISA CPUs, accelerators (GPUs, FPGAs) and high performance networks (IB, RoCE, Omnipath), making the management of heterogeneous clusters easy with a single software stack for different platforms.
Bare metal & VM provisioning: Implement your HPC solution on bare metal nodes or using an IaaS.
HPC resource manager & scheduler configured to achieve maximum efficiency in your cluster and have full control of workload execution.
Centralized Monitoring and Alerts thanks to a single intuitive control point which permits verification of the state of the cluster and receipt of alerts in case of faults.
The software stack includes the main open source tools for development and execution of high performance applications: compilers, scientific and digital libraries, MPIs, OpenMP.
On demand components
In addition to the functionalities present in the base configuration, Medooza can be enriched with a series of additional functionalities suited to the specific needs of each project:
- Optimized compilers and libraries
- Management of datacenter infrastructure
- High reliability
- Cloud burst
- Parallel/distributed/shared/object storage
HIGH PERFORMANCE COMPUTING