how to set budget limit on a compute instance created in azure ML studio

JA 131 Reputation points
2022-10-27T09:49:26.367+00:00

hi Team,
I want to set limitations on the VMs which are used for model training. Now, these compute instances are created using ML Studio and my requirement is to pause the model training if the pre-defined budget is over. This is to prevent user of unexpected charges and control the expenses.

But I am not able to figure out the way on how to do it.
1 way is to put a limit on the compute instance so that if it crosses the budget then its is paused and only upon resetting the budget, user can start it back.

To achieve this, I am thinking to set some actionGroup on budget to pause the VM or is there any automatic setting available to do this?

On top of everything, I am not able to find the filter which can put a budget on ML Compute. I can find till workspace but nowhere on compute part. How to control the model training budgets in that case?

Azure Machine Learning
Azure Machine Learning
An Azure machine learning service for building and deploying models.
2,418 questions
Azure Cost Management
Azure Cost Management
A Microsoft offering that enables tracking of cloud usage and expenditures for Azure and other cloud providers.
1,838 questions
{count} votes

1 answer

Sort by: Most helpful
  1. romungi-MSFT 41,321 Reputation points Microsoft Employee
    2022-10-27T14:36:24.467+00:00

    @JA I believe you can set budget for Azure ML compute using the following settings. I don't see this documented in detail in Azure ML docs but there is a section to manage costs using budgets.
    The following settings should work to set alerts on your compute usage within the workspace. I do not see an option to drill down on the compute type like compute instances, compute clusters or inference clusters though.

    The resource ID is what you will have to check to get the compute usage for your workspace. Just search with the workspace name for the Resource.Id filter value and you should be able to filter the cost of your workspace compute.

    254834-image.png

    Also, the compute instance offers a feature in the advanced settings to schedule the compute instance start, stop and auto shutdown. I have been using this feature to manage my compute instance in my workspace.

    Other ways to reduce costs related to Azure ML are documented here for reference.
    I hope this helps!!

    If an answer is helpful, please click on 130616-image.png or upvote 130671-image.png which might help other community members reading this thread.

    1 person found this answer helpful.