How does non-blocking I/O provide better throughput? The OS schedules the user process differently in the case of blocking and non-blocking I/O. When you block, the process “sleeps”, which leads to a context switch. When you use non-blocking sockets, this problem is avoided.
wayneseguin:
6 thins per slice over 8 slices can sustain 6 million hits per day with the Rails app we tested on (it did use memcached) from 4 distinct siegeing locations over the net.
wayneseguin:
I'd like to test with more sources
macournoyer:
that seems big to me, but can you compare that to something else?
macournoyer:
is it good?
wayneseguin:
Not entirely sure, that satisfied the clients needs
macournoyer:
ah ok
wayneseguin:
Plain mongrels were bombing out by that point
Extracting controller outside of rails using Rack on Thin