[Asterisk-Users] Re: FXO cards

[Rich Adamson]

> I know how to RTFM.  I have found no reference yet to how many x100P 
> cards I can run in one box.  Where did you find that reference?  I have 
> looked at the manual avaliable from *.org, downloaded the pdf of the 
> x100p card tech info, it doesn't say it anywhere.

You won't find it published anywhere. There has been some previous discussion
on this list relative to the topic, but I've not heard anyone truly
conduct any scientific evaluation to quantify the real limits.

Several people have made comments that each x100p requires a dedicated IRQ
as well. That's not true either. I'm running two x100p cards in production
and one of those cards is sharing the IRQ with three/four other devices
and one of those devices happens to be the asterisk ethernet port. The
quality of the two lines are identical (and good) once the echo problems
were addressed last month. One of the key issues seems to be that each 
x100p card "must" be able to properly handle approx 1,000 interrupts per 
second. If devices with low/no interrupt volume are combined with an 
x100p, there is not likely to be an issue. (That really does assume you 
know for a fact that other devices actually use the interrupts... many 
devices are tied to interrupts but don't actually use/generate them.)

The issue that I have not heard anyone try to quantify is more likely a
combination of processor speed, bus speed, application"s" running, etc,
and how much processing time is left to handle one/two/three/four 
times 1,000 interrupts / second without degrading/impacting other 
system functions and apps. (Don't think I'd care to try or publish that
test.)

So, the bottom line to how many x100p's can one have in a system is
"it depends", and my system loads/applications are not likely to be
the same as your's, or the next guy on the list. Therefore, no one is
going to be able to answer your question with any degree of accuracy.

I did find a reference in the archives (would guess from about a year ago)
where someone described a procedure to analyze the interrupt counts
every ten seconds. It was kind of handy for eliminating potential issues,
etc. (I don't have the reference any more.)

Rich