Transition clusters providing users with dedicated nodes to use shared nodes

[consideRatio]

Brief description

Many jupyterhub's setup by 2i2c were initially setup to start dedicated nodes for users, this has several drawbacks that we can avoid by scheduling multiple users to run on individual nodes.

Drawbacks of dedicated nodes

• inefficient use of cpu/memory paid for
  • increases cost
  • increases required cloud quota to request to scale to support X users
• increased average startup times (server startup, image download)

Rationale & value to our mission

• expected reduced cloud cost for communities
• faster startup times on average
• it becomes easier to handle events

Definition of completion and success

All communities use node sharing, even if options aren't presented like in many educational hubs.

Action points

• Overview all communities still setup to provide user dedicated nodes
• Help them transition

Related

• Let all clusters cloud infra support the instance types 4, 16, and 64 CPU highmem nodes #3256
• Add script to generate resource allocation (nodeshare) choices #3030