No capsules yet. Hit Run workload to spin one up.
Each customer gets a guaranteed, non-overlapping slice of the machine. Capsules are packed within a tenant's quota; the bars are live usage vs. the hard limit. A tenant can't borrow another's capacity.
Add / update a customer
A customer is a tenant with a guaranteed CPU + memory slice. Saving persists to the cluster config; admission then holds every workload to this limit.
Define a workload
Save a reusable task/service template. Launch it into its customer with one click — admission packs it within that customer's quota.
Pick a scenario — ember runs it and narrates each step, then jumps you to the right view. The demo explains itself.
Bin-packing & admission
Fill a customer to its CPU quota and watch ember refuse the overflow.
Tenant isolation
One customer hits its limit; another customer is completely unaffected.
Tasks vs. services
A task runs to completion and frees its slice; a service stays running, supervised.