Request for Information for Data Centre Co-Location Services

Tender description

Background to the Project / Contract: UCL currently has one main onsite data centre on its London campus and also makes use of co-location space hosted by a data centre provider located outside London. UCL wishes to appoint a Supplier for the provision of Data Centre Co-location Services, predominantly to host High Performance Computing (HPC) equipment. UCL is therefore looking to identify Suppliers who have the capacity and capability within the marketplace by way of seeking a response to this RFI. The requirement: UCL has identified the requirements for off-site data centre co-location space as a result of a number of current and proposed projects. These initiatives are coupled with a demand to provide more data centre space for customers. The interested Suppliers are requested to provide information for each of the following requirements summarising how they can meet and/or exceed UCL's requirements. HPC systems tend to require higher power densities than typical IT equipment. 30kW cabinets are now regarded as entry level, with 40kW and above likely to be needed in the near future. Due to the use of low latency network interconnects the cabinets need to be located closely together and ideally it needs to be possible to expand into adjacent space over time. Vendors of HPC solutions generally prefer to provide their own cabinets and these are usually delivered with systems pre-populated in them, which can present logistics challenges. There is an emerging requirement for innovative cooling solutions - notably the connection of servers to a water supply to provide CPU cooling. As the majority of HPC usage is for research activities, equipment is increasingly funded via Research grants to the university. This brings additional challenges as not all grant bids are successful. However, when they are, there are usually very aggressive timescales associated for purchase and installation of equipment. This means that UCL often needs to obtain data centre capacity at short notice, with a high degree of certainty that it will be available. Accurately predicting future capacity can also be very difficult.