IT Services: Supply and Installation of a High Performance Computer Facility

Descriptions

The University is seeking to provide a high performance compute (HPC) facility to support research and teaching. This will be vital resource for existing high-profile staff in many areas of current research and its importance is expected to grow in an era of expanding digital research and increasingly larger datasets, including computationally intensive approaches such as machine learning and generative artificial intelligence. The proposed solution should provide at least 15,000 x64 compatible compute cores at 2GB RAM per core, excluding log-in nodes and peripheral devices. An alternative configuration providing the same or better level of performance would be acceptable. The proposed solution should provide high-memory nodes containing at least one 4TB RAM and two 2TB RAM. An alternative configuration providing the same or better level of performance would be acceptable. The proposed solution should provide a number of GPU Nodes whose combined cost equates to 20% of the combined cost of requirements above. Data Storage The University requires the following storage tiers as a minimum: 1TB of scratch storage per node 100TB of fast scratch storage to service parallel tasks 2PB of fast shared scratch storage 1PB of archive storage Networking The proposal must include a low-latency 100Gb network fabric connecting all nodes. Extensibility The proposed solution must provide a seamless means to extend to make use of public and private cloud resources. Scalability The proposed solution must be able to scale for each use case (fast throughput parallel processing, single-node high core-count and high memory processing, and GPU workloads). Additional Support The responses should include an optional proposal for the provision of user support services. Please see ITT documents for full details.