Concurrency question

[maccopari]

Hi Christian. I was really impressed by this project. I was wondering how well this solution handles concurrency among different clients.. Lets say that 10 different clients tries to use the Face-Recogn on the same server, sending each one a different image. How will the server behave? It will be able to handle the prediction on each single image as a critical zone and ‘serialize’ each client requests?

[krystianity]

Hi @maccopari I have never load tested this solution. But as it is separated into multiple servers (at least one C++ tensorflow serving instance and 1 Node.js middleman) you should be able to scale this very well. In terms of concurrency there is no "shared" state between requests, therefore you can make as much concurrent requests (with different images of course) as your machine can handle.