asterisk users - May 2004 - Number of Digium cards in one box...

I know, I know, check the archives but I can't find an answer since the
new cards are well, NEW!

I understand the whole issue of expandability and flexibility of using a
T1 card and an Adtran 750. FXO or FXS, you mix and match.

With the new card offerings from Digium I can easily put a 4CO by 8
Station system together.  

Barring the extra interrupts, extra CPU Cycles of extra interfaces, is
there a reason why I would not do this??

How much overhead would the 4 port cards put on a system?? At what point
would the breakeven point be??

The cost of the T1 card is not the problem, the cost of a channel bank,
New, no-ebay, or used stuff here. I want to compare new apples to new
apples, not used.

Any comments???


Alex

----- Original Message ----- 
From: "Alex Lopez" <alex@opsys.com>
To: <asterisk-users@lists.digium.com>
Sent: Monday, May 03, 2004 5:37 PM
Subject: [Asterisk-Users] Number of Digium cards in one box...


How much overhead would the 4 port cards put on a system?? At what point
would the breakeven point be??

--------------

I have a deployed system with a 405, 2x40b, and a 31b.  3 spans on the 405
are in use, 1 to a provider, 1 in a p2p with another building, and 1 to a
750 with 6 fxs modules.

model name      : Intel(R) Pentium(R) 4 CPU 2.40GHz
stepping        : 5
cpu MHz         : 2394.206
cache size      : 512 KB

root@phone:~# uptime
 19:10:42  up 6 days, 12:07,  2 users,  load average: 0.78, 0.87, 0.88

  0:   56208454          XT-PIC  timer
  1:      10999          XT-PIC  keyboard
  2:          0          XT-PIC  cascade
  3:  562032946          XT-PIC  t4xxp
  5:  699933781          XT-PIC  Intel ICH5, wctdm
  8:          1          XT-PIC  rtc
  9:  562017411          XT-PIC  usb-uhci, wctdm
 10:  562030564          XT-PIC  usb-uhci, wctdm
 11:    7923921          XT-PIC  3ware Storage Controller, usb-uhci,
usb-uhci, eth0
 12:       3015          XT-PIC  PS/2 Mouse
 15:          7          XT-PIC  ide1
NMI:          0
ERR:          0

> I know, I know, check the archives but I can't find an answer since the
> new cards are well, NEW!
> 
> I understand the whole issue of expandability and flexibility of using a
> T1 card and an Adtran 750. FXO or FXS, you mix and match.
> 
> With the new card offerings from Digium I can easily put a 4CO by 8
> Station system together.  
> 
> Barring the extra interrupts, extra CPU Cycles of extra interfaces, is
> there a reason why I would not do this??
> 
> How much overhead would the 4 port cards put on a system?? At what point
> would the breakeven point be??
> 
> The cost of the T1 card is not the problem, the cost of a channel bank,
> New, no-ebay, or used stuff here. I want to compare new apples to new
> apples, not used.
The new TDM04B card (old card with new daughter boards) technical specs
seem to be somewhat of an unpublished thing.

System with two x100p's and one TDM04B shows:
[root@asterisk]# cat /proc/interrupts
           CPU0       
  0:    9953447          XT-PIC  timer
  1:        253          XT-PIC  keyboard
  2:          0          XT-PIC  cascade
  3:          0          XT-PIC  usb-uhci
  8:          1          XT-PIC  rtc
  9:  204875095          XT-PIC  ehci-hcd, eth0, Intel ICH4, wcfxo, wctdm
 10:          0          XT-PIC  usb-uhci
 11:   99034988          XT-PIC  usb-uhci, wcfxo
 12:       3451          XT-PIC  PS/2 Mouse
 14:     156855          XT-PIC  ide0
 15:     787553          XT-PIC  ide1
NMI:          0 
ERR:          0
[root@asterisk]# 

Don't see any interrupts associated with the wcfxs card. (Oh, I was
told by digium support the software routines for the FXO daughter
boards are in the wcfxs module. Looking at the file dates, that
appears to be correct. But, no interrupts. Cool!)

Don't have a clue what the limitations might be at this time, however
it would appear:
 a. echo issues with the FXO card are identical (if not worse) then
    the x100p's. Looks/feels like the x100p prior to about Nov 2003.
    Thirty seconds of decreasing echo after anwser.
 b. the FXO card spontanously thinks the pstn line is ringing, and
    executes the dialplan entry assoicated with that. Rings twice 
    and then disappears. (Note: callprogress=no/yes has absolutely
    no impact, as though it was not implemented on the FXO card.)
 c. CallerID seems to be a less reliable then the x100p. Nothing to
    back that up other then gut feeling.
 d. transmission levels are good and adjustable via rxgain/txgain,
    but has little impact on the stock echo problem from what I've
    seen thus far.

Based on informal comments, I'd have to guess the software necessary
to support the real FXO hardware needs are lagging by a fair amount. 
Still testing though...

Rich