Skip to content

Latest commit

 

History

History
81 lines (56 loc) · 3.41 KB

cost_model.md

File metadata and controls

81 lines (56 loc) · 3.41 KB
Raw

What Are Cost Models?

Configuring a cost model allows users to associate prices to metrics and usage, and fine-tune the costs of running their cloud. This is an essential piece of running OpenShift on Prem, because without a cost model we can generate any costs. Therefore, a majority of the features available for the cost models are related to OCP sources. However, you can still apply a markup or discount through cost models to our cloud providers (AWS, Azure, GCP, OCI).

Helpful Links:

  1. Cost Managements Docs:

OCP Cost Model Features:

Price List

A price list is a mapping used to associate prices to metrics and usage. We use this mapping on the backend to calculate additional costs.

Example

  1. Price list
    • usage metrics
    • monthly rates

Usage Metrics

The following usage metrics use a simple equation of usage * rate to populate an additional cost in the database:

cost_model_cpu_cost

  1. cpu_core_usage_per_hour
  2. cpu_core_request_per_hour
  3. cpu_core_effective_usage_per_hour

cost_model_memory_cost

  1. memory_gb_usage_per_hour
  2. memory_gb_request_per_hour
  3. memory_gb_effective_usage_per_hour

cost_model_volume_cost

  1. storage_gb_usage_per_month
  2. storage_gb_request_per_month

These metrics are multipled by the rate then aggregated to create the corresponding cost_model_{type}_cost column in the usage_costs.sql file.

After these columns are populated in the database, they are added to the overall cost in the provider map to generate the user facing cost.

Monthly Rates

Monthly rates are flat rates that are intended to cover a subscription cost.

Cluster and Node These cost are amortized, which means we distribute that cost evenly throughout the month so that each day has an equal portion of the total cost for the month. We distribute the rates applied to the node & cluster metrics within the monthly_cost_cluster_and_node.sql.

node_cost_per_month cluster_cost_per_month

using the following equation:

effective_usage / capacity * monthly_rate

These costs are also inserted into the cost_model_cpu_cost and cost_model_memory_cost; however, we insert a value into the monthly_cost_type column to distinguish these values.

cluster_cost_per_month

select * from reporting_ocpusagelineitem_daily_summary where monthly_cost_type='Cluster';

node_cost_per_month

select * from reporting_ocpusagelineitem_daily_summary where monthly_cost_type='Node';

PVC The monthly PVC cost follows a slightly different logic, where rate is divided by the number of unique persistent volume claims within monthly_cost_persistentvolumeclaim.sql.

Equation:

rate / pvc_count

pvc_cost_per_month

select * from reporting_ocpusagelineitem_daily_summary where monthly_cost_type='Node';