Support private clouds / On-prem infrastructure

[hassankhosseini]

Feature for: Larger enterprises who have on-prem or private clouds. These could include VMWare, OpenStack etc.

Currently Infracost shows prices for workloads running on public cloud providers. This feature would enable the central IT team or the infrastructure managers within larger companies to define a price book for their private cloud / On-prem infrastructure, and those prices could be used when providing cost estimates in pull requests.

The setup could be: we pull in different VM sizes with names, provide a way for prices to be input (cost per hour/minute/etc), and use those during estimation.

[maxime-sourdin]

In my company, we use a provider based on openstack

From what I know about the implementation used, billing is done in flavour, with obviously different names, and depending on whether the host is dedicated or not, same thing for reservation.

We also have network costs billed in a particular way, and specific billing for managed Kubernetes.

So the user could provide a json to parse, which would be taken in parameters like that:

Managed kubernetes: (price per cluster managed)
       Nodes (price per flavor)

Storage ( Price per Go)
    SFS:
         Type
    S3: 
        Type

Compute:
    Reserved instances:
          Flavor 1: : (price)
          Flavor xxxxx: : (price)
    Normal instances:
          Flavor 1: : (price)
          Flavor xxxxx: : (price)

Database:
        PostgreSQL:
          Flavor 1: : (price)
          Flavor xxxxx: : (price)
       mysql/MariaDB:
          Flavor 1: : (price)
          Flavor xxxxx: : (price)

Network (for all services): (price per go)
      Input (price)
      Output (price)

[sempervictus]

Far as OpenStack goes: might be feasible to query cloudkitty in environments which deploy it or pull usage information from the various telemetry sources/sinks about workloads.