Azure Machine Learning cost questions- When exactly does compute billing and stop?

Question

Azure Machine Learning cost questions- When exactly does compute billing and stop?

Seonho Kim 25

Azure Machine Learning cost questions

When exactly does compute billing and stop?

The moment you create a notebook? The moment you start the notebook?

If I create a notebook with GPU compute, do I have to pay the GPU compute price while using the notebook?

https://learn.microsoft.com/en-us/azure/machine-learning/tutorial-cloud-workstation?view=azureml-api-2#start-with-notebooks

Jennie Barrett 0 Reputation points

2024-05-28T07:42:47.3333333+00:00
In Azure Machine Learning, understanding the billing for compute usage is essential for managing costs effectively. Here is a breakdown of when compute billing starts and stops:

Compute Billing Start Time:

Compute billing typically starts when you initiate a compute instance or a compute cluster for training, scoring, or inference tasks in Azure Machine Learning.

For compute instances, billing usually begins as soon as the instance is provisioned and ready for use.

For compute clusters, billing typically starts when the cluster is started and available for running workloads.

Compute Billing Stop Time:
- Compute billing stops when you explicitly stop or delete the compute instance or compute cluster. - For compute instances, stopping the instance halts billing immediately, as you're no longer consuming resources. - For compute clusters, stopping or deleting the cluster terminates the compute nodes, and billing ceases for the stopped cluster.

Idle Time and Auto-scaling:

Azure Machine Learning offers features like auto-scaling, where the cluster scales up or down based on workload demand.

If auto-scaling is enabled, billing may continue for a period after the workload completes, as the cluster nodes may remain active to handle potential future tasks.

Idle time, during which the compute resources are provisioned but not actively used, may also incur billing charges.

Accepted answer

0 additional answers

Your answer

Jennie Barrett 0 Reputation points

2024-05-28T07:42:47.3333333+00:00

In Azure Machine Learning, understanding the billing for compute usage is essential for managing costs effectively. Here is a breakdown of when compute billing starts and stops:

Compute Billing Start Time:

Compute billing typically starts when you initiate a compute instance or a compute cluster for training, scoring, or inference tasks in Azure Machine Learning.

For compute instances, billing usually begins as soon as the instance is provisioned and ready for use.

For compute clusters, billing typically starts when the cluster is started and available for running workloads.

Compute Billing Stop Time:
- Compute billing stops when you explicitly stop or delete the compute instance or compute cluster. - For compute instances, stopping the instance halts billing immediately, as you're no longer consuming resources. - For compute clusters, stopping or deleting the cluster terminates the compute nodes, and billing ceases for the stopped cluster.

Idle Time and Auto-scaling:

Azure Machine Learning offers features like auto-scaling, where the cluster scales up or down based on workload demand.

If auto-scaling is enabled, billing may continue for a period after the workload completes, as the cluster nodes may remain active to handle potential future tasks.

Idle time, during which the compute resources are provisioned but not actively used, may also incur billing charges.

Answer 1

Hi Seonho Kim ,

Thank you for reaching out on Microsoft Q&A.

Great questions about Azure Machine Learning compute billing! Compute billing begins when you allocate or start a compute resource, such as a virtual machine or GPU cluster, which means charges start when the notebook is actively running and using these resources. Billing stops only when you explicitly stop or deallocate the compute resource, and simply closing the notebook will not halt billing. If your notebook uses a GPU cluster, you will incur GPU costs for the entire duration it is in use. To effectively manage costs, ensure you stop the compute instance in the Azure Machine Learning studio when not in use.

If you found this solution helpful, consider accepting it.

Share via

Azure Machine Learning cost questions- When exactly does compute billing and stop?

0 additional answers

Your answer