At the moment, users can choose to distribute costs based on CPU or memory usage

Some users have asked to do cost distribution based on some custom (Prometheus) metric generated by their application or e. g. by an API gateway.

We would need to tell the CMMO to gather those metrics and whether to charge a fixed amount per usage, or distribute the total cost based on the aggregate of that metric.

Use cases:

• Distribute the cost of this cluster to different internal customers based on some custom metric which includes the internal customer reference (NB from Pau: this looks more or less like a tag to me)
• Distribute the cost of this namespace, which includes 20 pods (one pod per internal customer, tagged) to 15 internal customers and then distribute the remaining 5 pods equally across all internal customers
• ...