[GLLUG] stupid question

[Adam McDougall]

On Sun, Dec 17, 2006 at 02:28:07PM -0500, eduardo cesconetto wrote:

  Here is a (stupid) question:
  
  Will it make any difference in speed to load applications(i.e.  
  Firefox) in the same window manager in different distros running on  
  the same machine? Is there a faster flavor oh *nix for older machines 
  (i.e. BSD or Linux or Unix)?
  
  []'s
  Eduardo
  
Many things can affect the application load time, some of them
you can try to make more consistant, while some you can only measure,
and some things are pretty hard to tell unless you figure out how to
single out as few differences as possible before benchmarking.

- free or quickly free-able memory available.  I consider buffer 
  and cache memory to almost count with free for this purpose.
- if swap is partially used when firefox is done loading, did the
  pc have to swap something out before firefox could load, or was
  it already swapped out?  In other words, did firefox get slowed
  down loading because the kernel had to make room in ram.
- additionally, if the computer has insufficient ram, differences
  in kernels, oses, distributions can affect how efficiently your
  computer handles the lack of ram.
- where the files (binaries, libraries, etc) are stored on disk
  - if scattered around HD, seek time increases load time
  - are files stored in more than one fragment, see above
  - position on HD (I think the beginning of the disk is faster)
  - you can try to make each OS more similar in this regard by 
    using the smallest reasonable partition sizes possible, and 
    perhaps by trying to make sure applications get loaded onto 
    the first parition on the disk.  There used to be some merit 
    in putting the swap partition first on the disk for optimum 
    swapping speed, assuming swapping was expected.
- whether or not other applications may have loaded libraries into 
  memory or disk cache already that firefox would use.  If I was
  purely concerned about testing firefox load time, I would test
  with a very simple window manager on each OS with as little else
  loaded as possible, and execute it as soon as possible after boot
  and only measure that first time since a reboot.  However, if 
  you wanted to test the load speed in a realistic user environment,
  you can do that comparison too.  
- file system speed, file system mount options, volume manager overhead?
- different dists may use different compile options or debugging
- different plugins or startup checks
- the OS itself can make a difference.  If you can try to normalize
  the other factors, a better comparison can be made.  But you can
  also choose to compare the out of box experience from the view of
  the average person who did an average install.
- kernel memory allocation algorithms
- kernel scheduler algorithms
- larger browser configurations will take longer to parse.  You could
  potentially slow down the browser startup time 
- exact firefox version
- probably more stuff I can't think of