Skip to main content

Plans, limits, and what they govern

You now know the four objects: session, cluster, workspace, organization. A plan bounds how much of them you can have.

This lesson covers the concept. The operational side (reading usage, controlling spend, choosing a tier for a team) lives in plans, usage, and billing, written for whoever owns that decision.

The one distinction worth learning now

A plan bounds two different things, and confusing them explains most of the surprises people hit.

Capacity is how much compute you can use in total over a billing period.

Concurrency is how much can run at the same time.

They fail differently:

Symptom
Capacity exhaustedWork is refused, or the bill is larger than expected
Concurrency exhaustedWork is accepted but queues, so sessions and clusters take a long time to start

The second is the one that gets misdiagnosed. A team well inside its capacity can hit concurrency limits constantly, and it does not feel like a limit. It feels like the platform is slow.

Why the distinction exists at all

Concurrency limits simultaneous resources because that is what the underlying machines actually constrain. Capacity limits accumulated resources because that is what gets billed. A plan has to bound both, or one of the two runs away.

Try this

Open your organization's dashboard and find the usage panel. Two questions:

  1. What is your capacity, and how much of it has this billing period used?
  2. How many sessions and clusters can run at once?

If you cannot answer the second from the dashboard, that is worth knowing before a team hits it.

Where to go for the rest

Everything operational is in the admin course:

Current tier details are in the platform documentation, which stays accurate as plans change. This course deliberately does not restate them.

Where that leaves you

You have the whole shape now: code in a session, work on a cluster, both inside a workspace, owned by an organization, bounded by a plan.

Choose your next step by role to go deeper on whichever part you actually need.