MX High Performance Compute Cluster Nodes and Interconnect

Tender description

The scope of this contract is for the supply, delivery and installation of a number of high-performance computing cluster nodes (HPC nodes) and associated low-latency high-bandwidth interconnect. The new HPC cluster nodes will be integrated into an existing cluster and storage infrastructure. The management of the cluster nodes will be based on the use of IPMI and CFEngine, with Redhat Enterprise Linux 7 as the operating system. Any hardware provided must support existing storage provided by Lustre and GPFS. All operating system software and any storage related clients required will be provided by Diamond. Lot 1: The scope of this contract is for the supply, delivery and installation of a number of high-performance computing cluster nodes (HPC nodes) and associated low-latency high-bandwidth interconnect. The total budget for this contract is GBP 250 000 exc. VAT. The new hardware procured as part of this contract will need to be delivered and installed by the 31 March 2020. The service will be required to be installed in such a manner to minimise the downtime of service to Diamond users as this new hardware will be physically replacing existing HPC nodes in use. The removal and disposal of existing hardware will be the responsibility of Diamond. The new HPC cluster nodes will be integrated into an existing cluster and storage infrastructure. The management of the cluster nodes will be based on the use of IPMI and CFEngine, with Redhat Enterprise Linux 7 as the operating system. Any hardware provided must support existing storage provided by Lustre and GPFS. All operating system software and any storage related clients required will be provided by Diamond. As part of the overall compute cluster environment at Diamond, the new HPC cluster nodes will be used to perform a wide variety of calculations from many small single core workloads to some (and increasing) use of MPI codes. Due to the need to integrate with existing storage infrastructure, which utilises a Mellanox Infiniband fabric, all of the compute cluster nodes must incorporate an appropriate Infiniband network card.