We're and ISP, and I got the following NAS:
* AS5300 with 120 MICA modems and 4 E1s
* AS5300 with 240 MICA modems and 8 E1s
* Lucent Max TNT with 96 modems and 3 E1s
All E1s are connected directly to our own Switch (ERICSSON) and we're using R2 signaling. In the last 4 months traffic is going down unusually and I'm investigating the cause. I started asking to the Switch guys to send me statistics per hour, of the last 30 days, of how many calls has been directed to the access servers and how many of them has been answered. (it should be 100%)
To my surprise the Lucent TNT is answering 100% percent of calls, meanwhile the AS5300 (both) are having a "no answer" rate between 2 to 15%... so I went to check the counters: and I have like 3% of calls in the "No answer" column... according to CISCO documentation, this is caused by high CPU load among other causes... so I checked the load in the last month (I got it graphed with MRTG/SNMP): the peak on the first AS5300 (the oldest one with 120 modems) was 10% CPU load and on the other AS5300 is 26%
On the other hand I'm sure that are free modems all the time, since I graph (with MRTG again) modems in use for each access server... for example:
AS5300 with 120 modems peaked at 60 in the last month
AS5300 with 240 modems peaked at 197 in the last month
As a side note, both access server got the latest IOS possible for each hardware config and the MICA portware is 2.7.4.0
Last week I enabled modem recovery commands this way:
modem country mica e1-default
modem recovery maintenance window 60
modem recovery maintenance action drop-call
modem recovery maintenance time 4:00
modem recovery threshold 5
modem recovery action download
...to see if that helps, but the counter still increase at the same rate. I also created a new modemcap to configure the modems with a recommended setting I found on CCO:
modemcap entry micav90:FD=&F:MSC=&D2S34=18000S40=10S54=456S10=50debugthismodemS71=4
On the E1 controllers, counters are always 0 (zero) errors... I never have seen an error.
According to CISCO documentation this calls never get answered by any modem and it should be under 1% of all calls.
But here is a strange thing: in each of the AS5300, I had configured the first 10 modems in a separate modem-pool, the only difference here is that they are configured with a different modemcap, to only negotiate V.34 (33.6k). This pool was made to service customers that have connections problems due to incompatible or bad client modems as a last resource. The weird thing is that only this modems have 0 "No answer" calls in their counters... the rest, which are configured for v.90 have at least 2% of the total calls in this condition.
I have followed all troubleshooting guides that I found, but I don't know what else I could do to debug the problem.... if someone has an idea, please reply...
I'll apreciate it.
Roberto Carlos Navas
