Large variation in processing time for the same document – Azure Document Intelligence (Standard tier)

Question

Hello,

We are using the Azure Document Intelligence service (Standard pricing tier) and we’ve observed significant variations in processing time for the same document when calling the API multiple times in a short period.

For example: Sometimes the same document is processed in 5 seconds, other times it takes 20 seconds or more, even though the document and request parameters are identical.

I am using the prebuilt-document model (Standard tier).

Could you please help me understand what factors may be causing this difference in processing times? I would like to explain to end users why the same document might take significantly longer (up to 4–5x) to process on different occasions.

Thank you in advance for your help.

Answer 1

Hi Daliborka Nedic

Resource Scaling and Multitenancy: Azure Document Intelligence operates as a multitenant service, which means that the processing times can vary based on overall demand and resource allocation at any given time. Variability in latency is common in microservices, especially during peak usage times.

1. Document Complexity: Different processing times can be influenced by document characteristics such as size and layout complexity. If the documents have high complexity (e.g., more pages, complex formatting), it might take longer to process.
2. Network Conditions: Fluctuations in network conditions can also affect processing times. The speed of your internet connection or any network congestion might introduce delays.
3. API Latency: Each call to the API can experience its own latency depending on server load, size of the incoming request, and other factors like backend processing capabilities at that moment.
4. Asynchronous Processing: Keep in mind that Document Intelligence uses asynchronous processing, meaning that there may be inherent delays that aren't necessarily tied to the documents themselves but rather how requests are queued and handled on the server side.

Here are some steps you might consider to reduce latency:

• Normalize File Sizes: Ensure document sizes are consistent and monitor how many pages you are sending per request.
• Check Azure Region Status: Keep an eye on the Azure region status to see if there are any reported issues or outages that could be affecting your service.
• Monitor Usage Patterns: Set baseline metrics for processing times based on typical usage to help identify when processing times deviate from the norm.
• Utilize Azure Monitor: Use Azure Monitor to track performance metrics and identify potential issues or trends that could be causing increased latency. SLA does not stay constant for Pay as you they need to load balanced with other regions in case they find slowness in one region. if they want dedicated availability they need to purchase commitment tier and load balance

Reference:

https://learn.microsoft.com/en-us/azure/ai-services/document-intelligence/concept/troubleshoot-latency?view=doc-intel-4.0.0

https://learn.microsoft.com/en-us/azure/ai-foundry/responsible-ai/document-intelligence/transparency-note?wt.mc_id=knowledgesearch_inproduct_azure-cxp-community-insider#system-performance

https://learn.microsoft.com/en-us/azure/ai-services/document-intelligence/concept/troubleshoot-latency?view=doc-intel-4.0.0#high-latency-or-low-performance

https://learn.microsoft.com/en-us/azure/ai-services/document-intelligence/how-to-guides/use-sdk-rest-api?view=doc-intel-4.0.0&pivots=programming-language-rest-api&tabs=windows#analyze-documents-and-get-results

I Hope this helps. Do let me know if you have any further queries.

Thank you!