Read this from another forum.
Does litespeed operate in the same manner?
Just to throw in a plug for nginx as a reverse proxy, we observe the following:
For every N dynamic requests per second that hit our application serves, our nginx proxies hold open 20N client connections (including long-running connections, keepalives, etc.).
nginx protects our application servers by caching inbound and outbound requests in memory or on disk before sending them on. Slow clients are a non-issue, and we're doing a lot of requests a second without stressing out nginx at all. Many thousands of keepalive connections at any give time too, speeding up our application for our clients.
For every N dynamic requests per second that hit our application serves, our nginx proxies hold open 20N client connections (including long-running connections, keepalives, etc.).
nginx protects our application servers by caching inbound and outbound requests in memory or on disk before sending them on. Slow clients are a non-issue, and we're doing a lot of requests a second without stressing out nginx at all. Many thousands of keepalive connections at any give time too, speeding up our application for our clients.