>If user requests are small and replies are large (e.g., most HTTP requests), DSR provides tremendous savings, because only a small fraction of traffic need traverse the load balancer. Even better, DSR does not require us to keep state on the load balancer device. See https://sre.google/sre-book/load-balancing-frontend/

Seems too good to be true. How can I implement DSR? Can I do so on EC2 / DigitalOcean, perhaps with floating ip addreses? Or, do I need physical hardware? Tried Googling, but there's very little info on how to actually implement DSR.