One of my developers is reporting that the connection rate to our Quova servers has become a bottleneck. He reports that his attempt to set the TCP keepalive option in our Ruby code doesn't seem to be honored.
I haven't yet looked at our code; but I'm curious if anyone here could offer advice, based on the vendor's sample Ruby request code on how to refactor it to make many questions over each connection (or otherwise eliminate/bypass the apparent bottleneck).
I'm painfully aware that I may be missing the right keyword or even whole concept here. RTFM is fine so long as you point me at T (right) F-ing M.
Solved - Quova's load balancer was configured to turn off keep-alive.