Wednesday, 20 October 2010

BT Infinity - More isn't necessarily more....

Another one of those moments - It just struck me when flicking through my inbox, BT is is rolling out Fibre to the Kerb across the UK, with slated speeds of 40Mbps down and upto 10Mbps up.

I had a good google around and couldn't find a clear indication of the contention ratio (down or up), or if there is QoS provision for example to offer Voice (over IP) services - interestingly the BT product includes a home hub, with internet voice, so I expect their to be QoS there.

VDSL based, so BT provide a VDSL modem with their service, and over WBC so other ISPs can wholesale provide the service.

Lets see how it works out. ... Alas I'm not on an enabled exchange - so I'm not going to find out for a while yet.

Tuesday, 19 October 2010

CIsco Dial-Peers and VAD - watchout for this one!

I've been using Cisco Gateways (2800s) for an Oracle Contact Centre PSTN connection. The agents started complaining of poor voice quality and clipping and missing words.

Silence suppression, I thought. But looking over the configuration (snippet below):

voice-port 0/0/0:15
 no vad
dial-peer voice 100 voip
 destination-pattern .T
 voice-class codec 1
 session protocol sipv2
 session target ipv4:
 session transport udp
 dtmf-relay rtp-nte
 no vad

All looked well with the world - VAD was turned off. ........... Well actually NO. I found out last night about the dreaded hidden default dial-peer which is used to match when other dial-peers don't - and guess what - this hidden, default dial-peer has vad enabled by default. The default dial-peer was being selected when Agent call legs and a outbound calls were being made from the IP side to the PSTN trunk.

 incoming called-number .

Added to the voip dial-peer above fixes this problem by catching all the VoIP to PSTN dialled numbers in the "." - Job done. Watch out for this little gem...

I also discovered that the CIsco 2811 operating as a PSTN gateway with VAD enabled stops sending RTP packets when I muted my handset, i.e. no RTP packets during silence. Now cool this saves bandwidth - but SUCKS!!! What's a dynamic (de-)jitter buffer going to make of that? Not to mention a NAT firewall?