I'm running a web server (apache, but I don't think there is really a difference), that serves a lot of requests to mobile devices (hundreds per second). The server is a mix of a pure web server with front end server for a J2EE app server (connecting through something like AJP).
What I notice is that I regularly have to increase the amount of Apache workers to keep up with the amount of requests. And when the bottleneck starts to happen, I see a large amount (~90%) of workers busy with "Writing to client" and "Closing connection".
I understand that this is normal. Considering that the clients are on a slower link, it will hold up Apache worker until the worker can deal away with an individual client request. But I was wondering if there is a dedicated system, h/w or s/w that can buffer the traffic on both request/and response. I've heard of all this "traffic acceleration" hardware, but I don't really know where to start looking for the one that will do just exactly that.
Thank you!