I have developed a rails application for my company.
It will be used by upto 3o people at a time. I have just read that
Mongrel will only process 1 request at a time (rails is only in 1
controller at a time). Does this mean that to get good performance i
need to run mongrel many times on different ports?
How many different instances of mongrel is recommended?
If this is true then it wouldn’t you agree it makes running mongrel a
lot less simple than “mongrel_rails start -pXXX”?
Also what is the simplest software to use that would allow me to run
multiple instances all accessible from the same URL? How does it work?
Do you assign 1 main port that all the other instances will be
accessible from?
if you only have 3 people accessing the application, i probably
wouldn’t worry unless you have a controller that takes a very long
time to process. most requests are handled within milliseconds, so
it’s really a non-issue.
if you do find yourself needing something more robust, take a look at
mongrel_cluster.
You’re correct: Rails can only handle one request at a time. HOWEVER,
you
have to know how long the reqest is. 30 people will most likey NOT be
using
the application at the SAME TIME because web apps connect you and then
disconnect you. In the unlikely event that requests overlap, the users
will
just wait in line until the server can get to them.
You should run performance tests and determine your site’s capabilities
in
terms of ‘requests per second’
Your deployment platform makes a huge difference. One instance of
Mongrel
running on Linux can often be as performant as three instances of
Mongrel on
Windows.
Use httperf on Linux to test your URLs to see how many requests per
second
you can handle with your current server.
As for simplicity:
Mongrel_cluster is a plugin that will let you set up multiple instances
of
Mongrel, each on its own port. You have to use Linux for this.
Pen is a simple load balancer that can link them together.
Say I have two instances of Mongrel running, one on port 4001 and one on
port 4002;
To start listening on port 4000 with Pen, I just do
pen -r 4000 localhost:4001 localhost:4002
General rules apply here too… you need to optimize your application.
Use
page caching whenever possible (like pages which don’t need
authentication)
and use action or fragment caching whenever you can to reduce the
request
time. Ensure you’re eager loading data when appropriate to avoid
unnecessary database access. Make sure you’re running in production
mode so
things run faster
And consult the community. We’ll help you out where we can.
General rules apply here too… you need to optimize your
application. Use page caching whenever possible (like pages which
don’t need authentication) and use action or fragment caching
whenever you can to reduce the request time. Ensure you’re eager
loading data when appropriate to avoid unnecessary database access.
With 30 users using it at once, even these steps are pre-
optimizations.
