[asterisk-users] Capacity for transcode G711 to G729

[RR]

Hi Matt,

> The best use I have seen is the newly converted IAX2 which can use
> multithreading in version 1.4, the beta of which should be released
> later this week.
>
> The best idea would be to compile Asterisk, run some tests (show
> translation recalc 60) with HT turned on, restart the box, bring it up
> with HT turned off and try again.

What's the best way to know for sure that you've everything setup the
right way to use HT with Asterisk? There're so many things, I'm not
quite sure if I am turning or conversely not turning enough things
on/off. I do the following right now:

- in /usr/src, I have the symlink of linux and linux-2.6 pointing to
the location of the src of the smp kernel like e.g.
/usr/src/kernels/2.6.9-34.0.1.EL-smp-i686
- Then do a fresh 'make' on asterisk with these symlinks in placehe
- Then reboot and turn on HT in the BIOS
- Then reboot with the smp kernel

Is that it?

If I compile with the linux/linux-2.6 symlink pointed to the kernel of
the NON-smp kernel, then reboot in the non-smp kernel but leave HT
turned on in the BIOS, does it matter? would that be enough? or should
i turn off HT in the BIOS as well to avoid it causing issues? In my
experiments with using (*) inside of a VM and doing SMP I'd seen that
simply booting into an smp kernel gave me timing issues even when (*)
was compiled against a non-smp kernel source. I don't see these
problems on a real machine but that's just one call. Who knows what'll
happen if I throw 100 at it.

Would love to see the results of this test you're setting up. At the
expense of bandwidth, maybe I'll just stick with g711 all the way
through and save money on g729 licenses and load on my machine. Any
thoughts on g726? Would using g726-32 be a good compromise on
bandwidth and cpu power instead of g711 or g729?

Thx
\R