Objective JH-7: Ephemeral Public Access Hub

[batpad]

Motivation

mybinder.org has been a phenomenal success in making interactive computation accessible to large swathes of folks. By simply clicking a link, without any login requirements, end users (and lay people) can explore computational narratives and notebooks, with environments (and contents) dynamically built as needed. It would be useful to experiment with a VEDA adjacent ephemeral instance.

Proposal

Deploy a mybinder.org-like public access (no authentication required) ephemeral hub (no persistent storage) with capped usage control (capped via total max amount of nodes available to be used at any given time).

Owner(s)

• @yuvipanda (2i2c)

Success Criteria

• [x] Ephemeral hub is set up at a particular URL
• [x] Total number of nodes that can be spun up for this is kept low
• [ ] Use cases for narrative storytelling with this public instance are explored

[batpad]

@yuvipanda - can we come up with a list of decisions that need to be made / questions that need answering here for us to proceed?

I'll start, but perhaps there's also a good check-list or so from the 2i2c side that we can work with @wildintellect to get filled / questions answered by the appropriate people.

• What should be the max amount of compute / memory we want to provide simultaneously? This would be a derivative of the max cap for infrastructure spend for this hub.
• For login, would we prefer:
  • No login at all
  • CI Logon or similar to have some login so we can do some user / usage tracking, but don't require further manual approvals
  • Other?

@yuvipanda I think if we can compile a list of questions that we would need specific answers for to go ahead with the ephemeral hub setup, @wildintellect can help us get answers so that we can go ahead here.

[wildintellect]

Requested timeline is to have this feature available before AGU in December. CC: @freitagb

[yuvipanda]

As a step 1, @batpad can you create an issue patterned around https://github.com/2i2c-org/infrastructure/issues/4246 (a template of sorts we use for new deployments like this) in the infrastructure repo?

I would say we start this with no login at all, and begin by capping this to max of 1 r5.xlarge node only (and we can go from there). Once you fill out the template there, i can add additional info about this node cap.

[smarru]

@yuvipanda whats your experience with mybinder on keeping the auth open with no login at all. Does it attract any intentional or unintentional abuse?

[yuvipanda]

@smarru https://github.com/cryptnono/cryptnono was the result of our experience handling mybinder.org. cryptomining is the primary negative problem, and cryptnono has helped address that reasonably effectively.

[wildintellect]

Do we plan to set a time limit?

[yuvipanda]

@wildintellect there's an idle timeout time limit already, and we can also set a 'max' timelimit of maybe 12h?

[yuvipanda]

@wildintellect you can check out https://binder.opensci.2i2c.cloud/ to see how this would look like.

[yuvipanda]

Scheduled to be set up in https://github.com/2i2c-org/infrastructure/issues/4576

[yuvipanda]

Has been set up at binder.openveda.cloud!

[batpad]

Re-opening the issue until end of PI:

This is done! But we have not done much publicity or documentation around it. @brianmfreitag @yuvipanda - now that we have this, would be great to discuss what we can do / make people aware, and discuss use-cases that we want out of this.