Monitoring dropped calls


#1

We occasionally get calls dropped in the middel of a conversation. Or we would call a spesific number and it would ring and just after it get answered it would drop it. We can call that same number from a cell phone without a problem. All our phones are on a seperate V-LAN - we do not use the switch on the Polycom phones, so the PC’s are not connected to the phone. We also use Cisco managed switches.

Is there anyway to track down (monitor) the problem with dropped calls? Do I have to turn debugging on, and if so which? Network monitoring tool etc.

Any help would be appreciated.

Thank you.


#2

post what hardware you’re using, along with your zapata.conf file contents…


#3

Hardware:

Single 3Ghz XEON CPU
512Mb RAM
TE110P
TDM04B
CAT 6 cabling
Cisco switches
Telco is Qwest T1 line signalling E&M wink

Zapata:
context=international
signalling=em_w
immediate=no
rxgain=2.00
txgain=0.50
echocancel=yes
echocancelwhenbridged=yes
echotraining=yes
callprogress=yes
;busydetect=yes
;busycount=4
callerid=asreceived
group=1
callgroup=1
pickupgroup=1
channel => 1-24

I was thinking to to try and turn callprogress off and enable busydetect and set the count to 8.


#4

Yes turn callprogress off! Also busy detect should not be needed as it is E&M so don’t set it either. Report back after turning callprogress off.


#5

yup, that is what i was going to suggest too…

also, what type of NIC are you using? if it’s an intel based chipset, you might try throwing a different manufacturer’s NIC in there - realtek makes a nice gig-E chipset that seems to be pretty popular. one of our servers was suffering from numerous (>10%) call drops, and putting a $30 NIC in helped immensely.

you might also try running zttest and report back on your results…if your card is missing interrupts, that could contribute to some drops.


#6

Sure I will report back - have to run it for a week or so like that. I was also thinking of disabling the onboard NIC (Intel) and slot in a NIC in a PCI slot. I do see some intterupts misses but on another machine that has no intterupt misses I get the dropped calls as well although it is around 10 dropped calls out of 2000 calls made per day.

What do you guys think is an acceptable % for dropped calls?

Added ZTTEST results:
Best: 100.000000 Worst: 99.98793 Average: 99.994881

That was after 62 passes.


#7

Those scores are fine.


#8

Can you also post the output of the CLI when a call gets dropped ?

It sounds to me like a special event on the protocol layer causing a call to drop, similar to a “statechange” or whatever your telco is signalling there.

It could help to know if your CLI is reading like “got hangup request” or “frame error, restarting channel” or something like this.