Increase Document Intelligence TPS Limit

Josh Gottesman 25 Reputation points
2025-06-21T00:07:57.37+00:00

We're developing a solution that relies heavily on Azure Document Intelligence for high-volume data extraction from large, scanned PDF documents. The default 15 TPS limit is causing significant sequential processing delays (e.g., processing a ~66-page document takes ~105s). This bottleneck directly impacts our operational efficiency and scalability. We are seeking guidance on the process and requirements for requesting an expedited TPS quota increase to meet our project's throughput demands.

Azure AI Document Intelligence
Azure AI Document Intelligence
An Azure service that turns documents into usable data. Previously known as Azure Form Recognizer.
2,121 questions
{count} votes

Accepted answer
  1. Vinodh247 34,741 Reputation points MVP Volunteer Moderator
    2025-06-21T11:40:31.8466667+00:00
    • Default TPS limit: 15 TPS per Azure region per Document Intelligence resource.
    • Implication: Even if your PDF is large, each page is a separate transaction; so, 66 pages = 66 transactions = ~4.4 seconds minimum, but with concurrency and latency overhead, it goes beyond 100 seconds if processed sequentially.

    Requesting a TPS Quota Increase

    Step-by-step Process:

    a. Go to Azure Portal

    • Navigate to: Azure Portal - Quotas.

    b. Select the Subscription and Region

    Choose the Azure subscription and region where Document Intelligence is deployed.

    c. Choose the Right Quota Category

    Under "Quota type", select Cognitive Services (or AI services, depending on portal view).

    Then choose "Document Intelligence (formerly Form Recognizer)".

    d. Request Quota Increase

    • Look for "TPS" quota.
    • Click Request Increase.
    • Enter the desired value (60 -100 TPS, depending on your scaling requirement).
    • In the justification, describe:
      • Document size (66 pages)
      • Volume (ex: N documents per day/hour)
      • Impact (ex: current processing time, business delays)
      • Architecture (mention batching/concurrency model)

    Alternative Options to Optimize Throughput

    Even with quota increases, consider:

    • Parallel document partitioning: If possible, split large PDFs into smaller segments and process in parallel.
    • Batch reading with Analyze API: Use the async versions of the Document Intelligence APIs to reduce client wait times.
    • Regional load balancing: If you have multi-region architecture, split workload across regions (East US, West Europe).

    Please 'Upvote'(Thumbs-up) and 'Accept' as answer if the reply was helpful. This will be benefitting other community members who face the same issue.

    1 person found this answer helpful.
    0 comments No comments

0 additional answers

Sort by: Most helpful

Your answer

Answers can be marked as Accepted Answers by the question author, which helps users to know the answer solved the author's problem.