On Sep 28, 2005, at 3:38 PM, Jeremy Dunck wrote:
On 9/28/05, Jacob Kaplan-Moss <[EMAIL PROTECTED]> wrote:
...In case you hadn't noticed, Django has a pretty large memory
footprint; between Apache, mod_python, Python, the database drivers,
etc., Apache server processes tend to weigh about 10M each (at least
on my servers).
I hadn't noticed. Do you just load up your web servers with memory,
or do you (allow Apache to) throttle concurrent requests?
Yes, and yes. Our current web servers have 2G of RAM, and our new
ones being ordered will have at least 4G. My philosophy has always
been that hardware is cheaper than programming time: 2G of RAM costs
about $100 these days, which corresponds to something around 2 hours
of programming time.
Does anyone have numbers on footprints for different servers? Any
pointers to comparisons of request/process/thread models between
various servers which support WSGI or mod_python?
(If my tone sounds shrill, I'm just curious, not upset.)
I don't have any hard numbers, but my impressions have been that
unless you need all the amazing stuff that Apache will do for you,
lighttpd+WSGI serves Django faster with a smaller memory footprint.
We will, however, be launching a new website in a month or so which
will most likely our first powered by lighty so hopefully at that
point I'll have some numbers to share; I'll certainly keep the list
posted.
Jacob
