The fully integrated platform designed for High Performance Computing combines preconfigured hardware, middleware and software tools that allow the user to perform HPC computations from day one, without any configuration burden. This “plug & play” appliance is flexible enough to be adapted to various needs of our customers. Moreover, the HPC Appliance is optimized to save energy, rack space and minimize total cost of ownership.
This M2DC appliance features a combination of high performance CPUs and GPUs that were carefully selected to maximize performance and energy efficiency. This complex distributed system is orchestrated by OpenStack which provides bare metal deployment of operating systems running on individual compute nodes. On top of that, the Slurm Workload Manager takes care of efficient allocation of hardware resources for your computational jobs. Finally, a set of pre-installed accelerated libraries for HPC, data analytics and deep learning saves your time so you can directly work on your domain problem.
Nodes connected with a high performance network allowing a superb scaling of HPC applications
Equipped with an embedded scheduling system that can offload the time consuming scheduling operations onto a dedicated hardware
MPI collectives accelerated by Message Passing, improving the performance of multi node MPI applications by reducing the latency of all-reduce operations that are common in HPC simulations
You can choose from a wide range of modules, which together form a heterogeneous cluster. These may be microservers with high-performance x86 or ARM64 as well as low-power ARM processors with the latter also providing GPU computing capabilities. But in addition to the CPU-based microservers, also FPGAs are available for specialized applications. PCIe extension cards are also supported for even further flexibility. The appliance nodes are connected with a high performance network that will allow a superb scaling of HPC applications. This is thanks to the high speed and low latency host to host RDMA communication over PCIe x8 which we call Communication SEE. Nodes will be connected with regular 10G Ethernet.
Flexibility allowing various configurations
High density architecture enabling integration of up to up to 27 high performance microservers and up to 3 PCIe accelerators
Preinstalled software enhancing the hardware capabilities
Intelligent power management optimizing the energy consumption
Greatly efficient and with lower cost.
9 High-Performance Microservers
48 Low-Power Microservers
Ready for the most demanding challenges.
27 High-Performance Microservers
144 Low-Power Microservers
The most powerful option.
45 High-Performance Microservers
240 Low-Power Microservers