Increase number of concurrent connections with apache server using redis

Question

I am running an Apache server on Amazon AWS with Redis (originally from here: https://github.com/RedwoodAdmin/RedwoodFramework). I am not at all well-versed in the back-end of this stuff (I am more of a data handler/econometrician using R/python/STATA) so apologize in advance if I ask the question poorly and without all the right details.

The problem we are having is that we are trying to run an experiment with 72 subjects concurrently (they are all in the same room, at a large computer lab at a university) by accessing the app/software through a browser (Chrome). When I tested the software simulating 72 subjects in 72 separate tabs, it runs fine. When we ran it on 72 machines, however, the software crashed/froze. I understand that different browsers/machines create different concurrent connections, when tabs do not. We had also run the software before successfully with 24 subjects at 24 computers. We then tested the software with 48 separate computers, and had the same freeze/crash problem.

I am hoping to get any ideas possible that could help us troubleshoot this issue.

I read before from here (http://stackoverflow.com/questions/3389496/how-do-you-increase-the-max-number-of-concurrent-connections-in-apache)

And then poked around the configuration files and found (in our configuration, probably installed ~2 years ago)

    $ cd /etc/apache2/mods-enabled

    $ ls -F

    $ vi mpm_event.conf

<IfModule mpm_event_module>

    StartServers                     2

    MinSpareThreads          25

    MaxSpareThreads          75

    ThreadLimit              64

    ThreadsPerChild          25

    MaxRequestWorkers         150

    MaxConnectionsPerChild   0

</IfModule>

In particular wondering if ThreadLimit parameter (or a combination of that or any of the others, min/max?) could explain what happened when we had trouble going from 24 to 48 (or more) separate workstations? Or it could be that we started with too few servers (2) making only 2*25 = 50 parallel connections available by default, and the server scale-up is too slow for an experiment like this? Though I haven't been able to try this out yet, and thought I'd take opinions here so we can try any number of things when we can reserve the lab again.

Again, I really apologize if the question doesn't have all the basic details you'd need -- please let me know what info could help diagnose this and I'll dig it up. Thanks very much.