Manage the TRE computing resources#
Since the Data Safe Haven platform runs on MS Azure cloud service, we have the ability to allocate computing resources as needed for projects, providing this comes within the project budget.
Typically resource management will involve adapting to project needs as per discussions with the project team; adding, removing or modifying the Secure Research Desktop (SRD) VMs associated with a given SRE.
Decision process#
How should we decide when/whether/how to increase compute resources?
At the project outset, the Project Team should have reviewed the project costing page and informed TRESA of the costing/compute requirements. This can be revisited at any point during the project, depending on the needs of the project.
If the Project Team informs TRESA that the computing resources of the TRE are inadequate for their purposes, or they have a more specific request for a particular Azure VM, do the following:
Discuss with the Project Team (e.g. the Project Manager) whether increased resources are needed and exactly what they would find most useful
Refer to the project costing page, but make sure to check the latest Microsoft Azure documentation for accurate costings
If a more expensive VM is required, determine whether this change should endure for the rest of the project, or a specific time window. If the latter, plan when would be appropriate to reduce costs by switching back to cheaper VM(s).
Follow the guides linked below to modify the computing resources as needed
Add comment(s) to the TRE GitHub issue reflecting decisions made and when changes were made in Azure, so there is a log of this
How to modify TRE computing resources#
There is info on resizing VMs (increasing/decreasing compute) in the DSH docs
Additionally using GPUs
As well as adding secure research desktops, which can vary in their VM configuration to requirement
The Data Study Group (DSG) TRE setup page has info on how we manage providing GPU-enabled desktops at short notice for DSG projects