Let’s say I have a system made up of a load balancer, which sits in front of multiple application servers, who in turn sit in front of a data store e.g. S3.
My system is for streaming videos which are stored in the S3 bucket.
So my client makes a connection to the load balancer saying “Hi please give me the next 10 seconds of this video”. The load balancer makes a request to the most appropriate application server which fetches the content. Now, when the application server is returning that content, is it returning it to the load balancer to pass back to the client, or is it returning it directly to the client?
It seems like sending it back through the load balancer creates unnecessary load, but I’ve never seen an architecture diagram where the application servers bypass the load balancer at any point either.