Voice Quality - Packet Loss

[Keith Joel]

Riddle me this, I've been troubleshooting a voice quality issue with a client recently and I'm curious to get some feedback from the community.  I've narrowed it down to packet loss between both the 2901 voice gateway (FXO for PSTN) and the phones (but also the 2960x switch and the phones), but I can't quite understand why.  Here's the situation:

Call Scenarios (media path perspective):

Inbound but WAN branch office answers = Router > 2960x Switch > WAN – Call is clear

Inbound and Local answers = Router > Switch > Phone – Call clips

Outbound from Local to PSTN = Phone > Switch > Router – Call clips, perhaps worse than inbound

Outbound from Local to WAN = Phone > Switch > Fortigate – Call is clear.

Ping Tests:

Cisco switch to phones – 10 to 15% packet loss

2901 VG router to phones (through switch) – 10 to 15% packet loss

2901 to Cisco switch – No loss

Cisco switch to HP switch, Fortigate, and End User – No Loss

HP and Fortigate to phones – No loss (through the Cisco switch)

It seems there's only loss between the router and phones or switch and phones, however no loss between phones and the WAN firewall or another switch in that room even though they have to traverse the 2960x as the phones are plugged in to that for PoE.  

Is this another chassis/rack grounding situation?

[Dennis Mink]

from what you are telling me, the issue is between switch and phones.

when you ping from cisco switch to phone, did you sample multiple phone?

did you ping within the same subnet,(the one the phones are in) or did you do and extended ping?

can you do a ping with a sweep range? see if it start dropping at a certain packet size.

did you test it putting a test switch in, with a test phone and see if that has a clean RTP stream? 

[Keith Joel]

Just to close off on this, TAC had pretty quickly dismissed the packet loss as something by design.  He referenced a DDoS issue from a few years ago.  It was still odd the phones responded 100% when from another vendors hardware even though those pings traversed the switch I encountered loss on, but at the same time maybe it's the best explanation.

The underlying voice quality issue was due to hair-pinning through CUCM (offsite) due to 'MTP Required' checkbox on the SIP trunk.  The provider had checked it rather than changing some of the handling settings on the sip profile.