Hi, On Tue, Jul 02, 2013 at 12:08:31PM -0700, Qingshan Xie wrote: > Hello Willy, > > > I am still unclear "how could 1 daemon HAProxy process handle thousands > requests/connections simultaneously or concurrently?" I thought the daemon > should fork children to handle connections, but I could not see any children > spawned when did a load-test with 100 concurrent users.
Fortunately it does not, we're in the 21th century now! Last I did something like this was in 1996 with my webroute project (haproxy's far ancestor). It's far too slow and you can't easily control resource usage when you proceed like this. > Could you help me to understand it? You should search on the net for "event driven programming", "multiplexing" and "asynchronous I/O". There is a lot of litterature on the subject and you'll probably find articles covering other products using the same principle such as nginx or squid. But I won't enter into a lesson here, the subject is too vast for this, it could take an entire book! Hoping this helps, Willy
