Skip to content

Latest commit

 

History

History
107 lines (77 loc) · 7.22 KB

service-limits.md

File metadata and controls

107 lines (77 loc) · 7.22 KB
title titleSuffix description services author manager ms.service ms.subservice ms.topic ms.date ms.author
Form Recognizer quotas and limits
Azure Applied AI Services
Quick reference, detailed description, and best practices on Azure Form Recognizer service Quotas and Limits
cognitive-services
vkurpad
nitinme
applied-ai-services
forms-recognizer
conceptual
02/15/2022
lajanuar

Form Recognizer service Quotas and Limits

This article contains a quick reference and the detailed description of Azure Form Recognizer service Quotas and Limits for all pricing tiers. It also contains some best practices to avoid request throttling.

For the usage with Form Recognizer SDK, Form Recognizer REST API, Form Recognizer Studio and Sample Labeling Tool.

Form Recognizer 2.1 Specific content goes here

Form Recognizer v3.0 specific content goes here [!INCLUDE]


Some more shared content here....

Quota Free (F0)1 Standard (S0)
Concurrent Request limit 1 15 (default value)
Adjustable No2 Yes2
Compose Model limit 5 100 (default value)
Custom neural model train 10 per month 10 per month
Adjustable No2 Yes2

1 For Free (F0) pricing tier see also monthly allowances at the pricing page. 2 See best practices, and adjustment instructions.

Detailed description, Quota adjustment, and best practices

Before requesting a quota increase (where applicable), ensure that it is necessary. Form Recognizer service uses autoscaling to bring the required computational resources in "on-demand" and at the same time to keep the customer costs low, deprovision unused resources by not maintaining an excessive amount of hardware capacity. Every time your application receives a Response Code 429 ("Too many requests") while your workload is within the defined limits (see Quotas and Limits quick reference) the most likely explanation is that the Service is scaling up to your demand and didn't reach the required scale yet, thus it doesn't immediately have enough resources to serve the request. This state is transient and shouldn't last long.

General best practices to mitigate throttling during autoscaling

To minimize issues related to throttling (Response Code 429), we recommend using the following techniques:

  • Implement retry logic in your application
  • Avoid sharp changes in the workload. Increase the workload gradually
    Example. Your application is using Form Recognizer and your current workload is 10 TPS (transactions per second). The next second you increase the load to 40 TPS (that is four times more). The Service immediately starts scaling up to fulfill the new load, but likely it will not be able to do it within a second, so some of the requests will get Response Code 429.

The next sections describe specific cases of adjusting quotas. Jump to Form Recognizer: increasing concurrent request limit

Increasing transactions per second request limit

By default the number of concurrent requests is limited to 15 transactions per second for a Form Recognizer resource. For the Standard pricing tier, this amount can be increased. Before submitting the request, ensure you're familiar with the material in this section and aware of these best practices.

Increasing the Concurrent Request limit does not directly affect your costs. Form Recognizer service uses "Pay only for what you use" model. The limit defines how high the Service may scale before it starts throttle your requests.

Existing value of Concurrent Request limit parameter is not visible via Azure portal, Command-Line tools, or API requests. To verify the existing value, create an Azure Support Request.

Have the required information ready:

  • Form Recognizer Resource ID

  • Region

  • How to get information (Base model):

    • Go to Azure portal
    • Select the Form Recognizer Resource for which you would like to increase the transaction limit
    • Select Properties (Resource Management group)
    • Copy and save the values of the following fields:
      • Resource ID
      • Location (your endpoint Region)

Create and submit support request

Initiate the increase of transactions per second(TPS) limit for your resource by submitting the Support Request:

  • Ensure you have the required information
  • Go to Azure portal
  • Select the Form Recognizer Resource for which you would like to increase the TPS limit
  • Select New support request (Support + troubleshooting group)
  • A new window will appear with auto-populated information about your Azure Subscription and Azure Resource
  • Enter Summary (like "Increase Form Recognizer TPS limit")
  • In Problem type select "Quota or usage validation"
  • Select Next: Solutions
  • Proceed further with the request creation
  • Under the Details tab enters the following in the Description field:
    • a note, that the request is about Form Recognizer quota.
    • Provide a TPS expectation you would like to scale to meet.
    • Azure resource information you collected.
    • Complete entering the required information and select Create button in Review + create tab
    • Note the support request number in Azure portal notifications. You'll be contacted shortly for further processing

Example of a workload pattern best practice

This example presents the approach we recommend following to mitigate possible request throttling due to Autoscaling being in progress. It isn't an "exact recipe", but merely a template we invite to follow and adjust as necessary.

Let us suppose that a Form Recognizer resource has the default limit set. Start the workload to submit your analyze requests. If you find that you're seeing frequent throttling with response code 429, start by backing off on the GET analyze response request and retry using the 2-3-5-8 pattern. In general it's recommended that you not call the get analyze response more than once every 2 seconds for a corresponding POST request.

If you find that you're being throttled on the number of POST requests for documents being submitted, consider adding a delay between the requests. If your workload requires a higher degree of concurrent processing, you'll then need to create a support request to increase your service limits on transactions per second.

Generally, it's highly recommended to test the workload and the workload patterns before going to production.