Form Recognizer Load Balancing

Question

Few questions regarding load balancing with the Form Recognizer service. I would be curious to know the answer for the on premise container, as well as the web container.

How many documents can each docker instance handle? Would multiple instances have to be set up in order to run documents efficiently?

• if scaling to multiple machines is needed, is there a way to uniquely identify which docker instance the POST request for analysis was sent from

My use case would be potentially handling several thousand documents a day.

Answer 1

Thanks for reaching out. Please review the document for the minimal requirements. Based on our testing, the throughput could be about 6 docs/minute (it also depends on many other factors, like document type, pdf, image, and size). One instance would be sufficient (if it meets the minimum requirement). In Post command, you can not specify a docker. Hope this helps!