Costs
Costs can help you find out which projects are the most expensive on your Conveyor installation. Using that information, you can target these and make them more efficient.
Supported Costs
Currently, costs are calculated for:
- Base cost for running a cluster
- Batch applications scheduled with the new (V2) Conveyor operators: ConveyorContainerOperatorV2 and ConveyorSparkSubmitOperatorV2
- Streaming applications
- IDE sessions
We do not calculate the costs of:
- The Airflow scheduler and web application running in an environment
- Batch applications scheduled using the legacy (V1) Conveyor operators
- Notebook sessions
How costs are calculated
The calculated costs are an estimation, we do not have visibility into extra discounts you might have on your cloud account from an AWS Savings plan or by using Reserved Instances.
We only take the instance costs of your jobs into account. For most jobs, these should represent the majority of the running cost of your application. Depending on the nature of your application, some jobs might incur significant networking or storage costs (f.e. S3, Blob, ...) as well.
Costs are calculated from when a container starts running until it is stopped. This duration is then multiplied by the cost of the machine used to run the application. The Conveyor instance you specified is used and matched with the price of the machine used, divided by how many jobs of that conveyor instance could run on the machine.
For example when launching a job with the mx.micro type, the mx.micro type runs
on an m6i.xlarge instance which costs about $0.214 per hour in the Ireland region.
On this machine type 16 mx.micro instances can be fitted,
meaning the job will cost $0.214 / 16 ≈ $0.0134
When using spot instances, the average daily spot price for each instance type is used, instead of applying it on a per-minute, per-instance basis. This is a simplification of how costs are calculated by the cloud providers, but it provides a good enough estimate in practice.