HPC for production and intensive calculation ...as a service




Ready-to-go tools



We have integrated a whole range of ready-to-go calculation tools. Simply select the code or the application to be launched, whether you want to run calculations or work on tools with a GUI. We strive to make as many tools available as possible, directly from our interface. There is nothing to install on your workstation as everything is launched from our portal. Do you create your own code? No problem, you can also install and compile it in your working environment. If you want, we can then help you make it available to the scientific community, by integrating it into our portal.




A set of adapted networks



At OVH, our HPC cross-connection is based on a 10 Gbps low latency Ethernet network for MPI communications between servers. The flexibility of managing Ethernet networks is linked to the performance of the new switches, which have comparable latency to Infiniband switches. Finally, so that you can work remotely on our tools, OVH has deployed its own network of 33 peering points spread across the world, so whether you're sending and retrieving your data, or working interactively, we guarantee you will have privileged access regardless of your location.




Flexible subscription models



Through our offers, we aim to democratise HPC, which is why we have set up access policies with maximum flexibility, in addition to offering excellent prices. As computing resource requirements vary from one business to the next, and even from one moment to the next, depending on the stage in the project cycle (R&D, development, testing, support), you can really reserve dedicated resources or usable hours of calculation on one or multiple servers simultaneously.




Physical servers designed for computing



In contrast to cloud solutions, OVH HPC infrastructures are made up of physical servers, which make up the computing nodes. This is a substantial advantage over infrastructures based on virtual machines, which actually slow down the calculations with their virtualisation layers and sensibility to the overall load of the master server. With physical servers, performance reproducibility is guaranteed.



A low latency latency, high performance network



One of the characteristics of HPC is the almost systematic need for a highly efficient network between the different computing servers of the cluster. This is to maximise bandwidth and decrease latency time so as to accelerate the data exchanges, which are usually large during calculation.


That's why our servers are inter-connected via a 10 Gbps low latency Ethernet network as standard, which can be substituted for an Infiniband 40 Gbps network for maximum performance.



Complete solutions for working from home



Calculation is great, but exploiting the results of your calculations is even better. You can access your data remotely using our KVM technology, which has been designed for scientific visualisation. Together with our web interface, it enables you to launch all the tools you need to work directly. No matter where you are, you can access your data and applications without having to download your results locally.