what do you guys use for monitoring call quality in production?

ciscog33k

I'm curious what tools people use to keep tabs on MOS etc. I've used VQmanager in production and thought it was really awesome for sip debugging (especially since asterisk is so lacking in that area), but the jitter and delay metrics never seemed right. In retrospect, I think there's a strong chance that it was because I was running ntp from a VM and at the time didn't know any better. I was mostly using it to aid in sip debugging so I never worried too much about it.

azaghul

Can't speak for production, but I've tried PRTG Network Monitor in the lab. It measures MOS, ICPIF, delay and jitter.

ciscog33k

I was just reading the docs and PRTG need sensors at both endpoints or cisco SLAs (which i haven't studied but that I'm assuming operate between cisco voice routers), so it doesn't work if one of your endpoints is a trunk provider on the internet.

What you do with VQmanager is configure SPAN and have it listen to all voip traffic. Then it pulls the RTP Sender/Receiver reports and computes the values it needs for the metrics. This method obviously depends on timestamps so both endpoints need to have very accurate time. Since being off by even a little bit can really mess up your jitter and delay stats.

Aside from that, I can see live calls, click on them individually, and then see all the sip packets that have been sent back and forth for that call, the ports being used, etc. I could run queries on thousands of calls to pick out errors, etc. Since asterisk's built-in troubleshooting options are virtually non-existent, it was hugely helpful for looking at what was going on in a system putting through a large volume of calls per hour.

azaghul

You can setup agents on PC's at each location with PRTG, but you can also configure IP SLA's on each site router and configure UDP jitter. It'll give all the values (and probably more) I mentioned earlier. But that still won't help with MOS values to your provider. If they use Cisco gear, they should be willing to setup what you need, if not they should be able to provide some reporting to you. My previous WAN provider was willing to configure SNMP and NetFlow for me, it also meant I could keep an eye on the network and keep them honest.

The version of PRTG I tested lost the decimal place, so a MOS of 4.1 was displayed as 41. It's also strick on how it counts sensors, so they can add up quickly.

Testing VQManager is on my to do list.