High-street bank Abbey, a pioneer in IP telephony, has
begun implementing service level monitoring to ensure that IP-based
calls are of the same quality as traditional analogue
calls.
IP telephony is particularly susceptible to problems such as poor
sound quality and failures, said analyst group Gartner, since it is
a real-time application that relies on a shared network
infrastructure,
Gartner said that although Multiprotocol Label Switching networks
can prioritise time-sensitive data such as voice traffic, the
technical measurements used to derive quality of service - such as
packet loss, jitter and delay - are not sufficient to determine
whether the quality of a voice call is acceptable to a user.
Instead Nigel Chisnall, technology strategist at Abbey National,
has used a technique known as Mean Opinion Score to measure the
quality of voice calls on the bank's IP network.
This is a benchmark for measuring quality of telephone calls, based
on surveying samples of users on the quality of recorded phone
conversations.
To set a benchmark on which to assess the quality of calls on the
IP network, Chisnall checked call quality on a five-mile copper
link across landlines in Ipswich. This gave a Mos score of
4.6.
Using voice quality assessment technology from Psytechnics, which
specialises in monitoring technology, Abbey's VoIP provider, BT,
implemented and maintained the Mos-based service-level agreement
for voice quality at the bank.
Abbey's SLA will require VoIP voice quality to be comparable to
landline quality across the entire network of 750 offices - a total
of 9,000 IP handsets.
At a Burton Group conference earlier this month, Chisnall said he
chose a Mos score of 3.7 and tested the network in Durham, central
London and the Isle of Wight to check call quality on the IP
network.
He said Psytechnics' non-intrusive technology supports continuous
monitoring. "We have 400,000 calls a week and needed to check all
of them," he said. "Psytechnics provides us with early diagnosis of
a problem." This is used by the bank to predict where likely
problems will occur.