Workspace User and Shared Storage

[marrobi]

• Provide User Storage to Researchers across Windows and Linux VMs #1566

• Shared storage in the workspace #1266

• Specify the storage quota for workspace shared storage #1267

Notes

• User storage does not need to be private, but dedicated to the user. Private per user storage may come later.
• Drives should be mounted into relevant locations - the mounting is a function of the user resource template - it may be different for different user resources.
• Creation of shared storage will be a function of the base workspace template
• Creation of user storage - TBD - but possible when a user resource is deployed to create if does not exist
• User shares should be created per user, with a unique identifiers - such as Azure AD object ID.

[marrobi]

Questions:
Should storage be added automatically or do users need to run a script retrieved from a portal?
Should users be able to add new storage (is this just a workspace service?) and connect to their machines via script?

[yorkecp]

Shared storage should be seen as the preferred location for users to store artefacts which they are creating and working on. This gives a degree of protection where artefacts are not lost if a virtual machine is removed.

As such it should be easily accessible by the user and so I think it needs to be automatically provisioned when a user is either added to a Workspace or when they first access it.

Within a workspace I would have no requirement for private space but a structured filesystem with for example:

users\user1
users\user2
shared
sourcedata

where sourcedata is a read-only "copy" of the source data ingested into the TRE through some form of secure data ingestion process.

[yorkecp]

Question:
Should there be controls over the size of the storage. This may be a cost issues which could be managed by #450 but a simple max size may be useful.

[marrobi]

@yorkecp what about the fact users often store files in their "home" directories by default.

We have previously redirected that to per user shares.

However your shared only model above does simplify initial requirements.

Adding a quota to the shared storage at deployment time should not be an issue.

[yorkecp]

Yes if home directory mapping was done that would drive compliance and make it even easier for users to fall into the pit of success. I was looking for a simple solution in the first instance.

Mapping home directories across multiple OS instances may be more of a challenge given our requirements around both Windows and Linux environments.