The RSS compute and storage infrastructure underpins all computational and data based research activities in UoE. All infrastructure hardware has a finite defined maintenance lifetime. A typical supported lifetime for server and storage hardware is 5 years (the majority of vendors will not supply maintenance for longer periods). University of Edinburgh research infrastructures have been operating in a “rolling refresh” model for the last 5 years. This means regular annual investment to keep infrastructure operating rather than large scale complete replacement every 5 years or so.
To maintain a 5 year replacement cycle for the EDDIE, DataStore and other infrastructure approximately 20% of capacity must be replaced annually. Similarly for the high performance storage supporting EDDIE 20% has to be replaced annually to maintain the maintenance cycle. These two actions are linked in that the HPC storage has to be closely matched to the compute infrastructure to maintain IOPs performance (typically spindle count).
Due to budgetary and staff resource limitations in. 20/21, some of the regular replacement/refresh that would have been expected to occur in that period was postponed. In order to maintain operation of these infrastructures this hiatus must now be recovered.
A managed infrastructure replacement cycle provides a managed approach to investment and cost management. Sustained investment in computational and data infrastructure underpins UoE competitiveness (as outlined in the University Strategy).
This project will deliver the necessary upgrades to the compute and storage infrastructure. Additional project documentation can be found in the project SharePoint area.
Current project status
|Report Date||RAG||Budget||Effort Completed||Effort to complete|
|May 2022||GREEN||80.0 days||54.0 days||35.0|