High performance computing (HPC) infrastructures are made up of a variety of technology hardware and software products built by various manufacturers, and as such can be complex to interact with efficiently. Mellanox Technologies provides an interfacing and management system that ties all these parts together as a whole, and then allows the system to be “partitioned” into sections. These sections allow different users to access the power of the whole system without interference or even the knowledge of other users. But how do you prove such a system works, without a HPC infrastructure of your own to test it on?
Mellanox Technologies’ own computing infrastructure is restricted, so they require access to a more advanced supercomputing system when comes to benchmarking, testing new applications or demonstrating to customers that their technology works on a larger scale. The Hartree Centre provides the supplier with flexible access to its supercomputing capabilities for benchmarking and testing purposes through Hartree partners OCF, who manage the HPC on-demand enCORE service.
The collaboration between Mellanox and the Hartree Centre is bringing mutual benefits to both organisations. The Hartree Centre uses Mellanox’s technology to provide a service that is partitioned off, enabling its customers to keep their data private when working simultaneously, allowing a bigger throughput of projects as many clients can use the supercomputing environment at once. This system can also help to identify inefficiencies or problems within the systems, so solutions can be found and dealt with efficiently. For Mellanox, the Hartree Centre enables them to access millions of pounds worth of some of the most advanced supercomputing equipment in the UK, without the impossibly high cost of building their own larger-scale, in-house infrastructure. As well as the potential to use the Hartree systems for R&D testing and benchmarking purposes, this allows the company to prove its technology to potential customers in a real-life case, enabling the company to realise new commercial opportunities and secure future revenue.
“Our customers typically build large-scale compute clusters, whom during the new cluster
bidding process, require real-world application performance benchmarks which preferably
are run on clusters similar to or as close as possible in size. Through our collaboration with
the STFC Hartree Centre, we are able to provide customers in multiple market areas with
the performance characteristics they need to reach an informed decision.”