 Julian Anastasov <ja@xxxxxx> wrote:
> 16,based on our calculations we needed to bump it to
> 26 to handle what were will be throwing at it.
>
> It is insane to use 26. That means 2^26 * space
> 2 pointers. On x86 it takes 512MB just for
> 2^POWER = AVERAGE_NUM_CONNS/10
>
> The magic value 10 in this case is the average
number of
> conns expected in one dlinked list, the lookup is
slower
> for more conns.
>
> Example:
> POWER=16 => 65536 rows => 655360 conns, ~10 on each
> row
>
Interesting... the help for that entry in the kernel
was misleading to me then. What I got out of it was
that each connection was each entry is 128+8 bytes,
and you should calculate the max connections you
expect to get times the number of seconds the
connection will last in the table (600 seconds
persistance set for port 443 currently), and that was
how much memory you needed, so I calculated the value
of X to get me to that number (2**X).
Max connections we can handle at once is 480.
480*600*136=approx 39meg.
2**26 is 64m
I'll read through that section of the HOWTO.
I'm not sure this will resolve the network issue we
are experiencing though, since we never made that
change to the backup director.
Thanks,
Dan.
__________________________________________________
Do You Yahoo!?
LAUNCH  Your Yahoo! Music Experience
http://launch.yahoo.com
