A cluster of your own, for as long as the work takes.

Ask for the machine your job needs. It is built on hardware that can actually run it, it is yours alone while you have it, and it goes back when you are done — so the next person can have the hardware.

Sign in Request access

Yours, not shared

Your own scheduler, your own network, your own login node. Nobody else's job queues behind yours and nobody else sees your data.

The hardware you asked for

Say you need GPUs or a fast fabric and the cluster is placed on a machine that has them. If nothing can satisfy the request, you are told why rather than handed something that will not run.

Software that travels

Applications arrive as signed containers and run unprivileged, so what you test is what runs on every node.

A browser is enough

Submit jobs, watch them, open a shell or a desktop from the web portal. Plain SSH still works if you prefer it.

How it works

  1. Describe the workHow many nodes, how much memory, which applications, and how long you need them.
  2. It is placedThe control plane finds hardware that meets the requirement and reserves it, so two requests cannot be promised the same card.
  3. It is builtScheduler, shared storage, applications and the portal, ready to use.
  4. It is returnedAt the end of the lease the cluster is released and the hardware goes back into the pool. Ask for more time before then and the lease is extended.