Brian Candler
6/2/2007 1:00:00 PM
On Sat, Jun 02, 2007 at 06:00:43PM +0900, Robert Klemme wrote:
> >Earle Clubb wrote:
> >>I have a setup where cron is calling N instances of a particular Ruby
> >>script simultaneously. It seems that as N increases, the delay before
> >>each script executes its first line of code increases by a little less
> >>than N seconds. I'm not sure if this is an issue with cron or if it's
> >>the cumulative startup time for N ruby interpreters. An ideas on how to
> >>speed this up to a near constant (~<1sec) delay? Is there a way to run
> >>all N scripts simultaneously with a single interpreter process? Thanks.
Well, you can't make a CPU run faster than itself. So let's say for the sake
of argument Ruby has a one-second startup time, and this is a CPU
limitation. Then if you run five Ruby scripts in five separate interpreters,
it's going to take 5 CPU-seconds of CPU time to start them all. If they all
share the CPU time equally, then it'll be shared 5 ways, so it will take
each of them 5 seconds of real time to execute their 1 CPU-second of work.
If the startup overhead is a problem to you, you could just try running all
the scripts one after the other within the same Ruby interpreter:
#!/usr/bin/ruby
# allscripts.rb
require 'script1.rb'
require 'script2.rb'
require 'script3.rb'
require 'script4.rb'
require 'script5.rb'
This may work as long as they are written in such a way that this doesn't
matter (e.g. they don't pollute the interpreter so badly that the following
one doesn't run, or depend on atexit {} or the like to function)
Alternatively, you write a single ruby process which starts and then:
(1) forks off a new child for each script, or
(2) starts each script in a new thread.
Brian.