Has anyone done any stress-tests which is more responsive under larger web traffic, im talking about simultaneous users load? Seen couple threads people asking testing their solutions on web but fairly nothing about results of these tests. Is standalone capable serving similar loads as cgi+apache solutions?
Stand alone should be able to handle just as many with better responsiveness and a lower load on the server. The Apache CGI layer adds overhead to the process which is why stand alone is faster. Apache does not launch multiple WE instances to handle the load, so any WE limitations are going to be the same.
You can run multiple stand alone WE instances behind something like nginx for load balancing. A rough guess is that one WE instance per CPU core on a server would be the most effective configuration, assuming you are not bound by RAM or disk I/O. I would say test this though.
Note: for WE CGI on a Windows Apache server, you need to install mod_perl. Without it the CPU load will shoot through the roof, and WE responsiveness drops like a rock with just a couple of users. I haven't fully investigated this on the other OSes.
Can you change the distribution methods if needed just by build settings, so no code changes are necessary?
On the WE side, yes. On the server if you want the URL to remain the same then of course you need to take additional steps.
I've found that the maximum simultaneous connections from a single client machine is 5. That seems to be a client side limitation and not a WE limitation. It's also by OS, i.e. my Mac and one of my Windows VMs can hit a WE app with 5 connections each.
I have 4 machines at my disposal, 5 including a VM, so I can hit a WE app with 25 simultaneous connections. In testing I've found that this slows initial launch a bit, but not dramatically, i.e. I would definitely say WE can handle more than 25 users at once on a dedicated server or a solid VPS.