In the first of two case studies, Helen Beckett looks at how the distributed nature of a voice over IP phone system adds to its resilience in the face of unexpected incidents.
When Brunel University began migration to a voice and data network, it little realised that the case for voice telephony would be proved in such a dramatic way.
A major incident occurred just days before roll-out was completed. It demonstrated the network's resilience and gave an early glimpse of the kind of flexibility that IP phone telephony could bring to a business operation.
Twenty-four months into the phased IP telephony project, a central heating valve that heated the entire Uxbridge campus through underground ducts blew. The valve was adjacent to the administration block that also housed the central - and still analogue - PABX.
The temperature in the PABX room rose to sauna levels and then fell within a few minutes. The team would later realise that the PABX cards were destroyed as all the circuit boards were exposed to 100-degree temperature and high humidity.
At the time though, any thoughts of a rebuild were quickly replaced by a call for more immediate action. The main switchboard was down and the university cut off from the rest of the world. The block housed not only the vice chancellor, but finance and student registration staff too, and 350 employees were without phones.
"It was the day before payday, so it was vital we got critical operations up and running as soon as possible," recalls Simon Furber, network manager.
Furber was away in Miami at the time, and so the disaster recovery operation fell chiefly to Will Templeton, IP consultant with his team. A call to BT's provisioning centre to reroute the main switchboard number to an IP line into the campus immediately created a host of contingency options. Templeton opted to create a user profile, login and hunting group from Cisco Call Manager.
The biggest obstacle to recovery was physically gaining access to the damaged administration block, which had to be dried out for 48 hours. In the interim, key staff members in the finance and administration departments were assigned IP phones and relocated to a small disaster recovery space abutting the IT room.
The administration staff had been the last to be switched to the new network, and so had not yet received training in VoIP. But Templeton and his team created user profiles in batches for administration staff and uploaded them to the Call Manager server. The basic profile consisted of the ability to call in and out of the network for national and mobile calls. The administration team was back in action in two days.
Had the new voice and data network not been installed and ready to plug into, and had the team been reliant on the legacy analogue system, then recovery would have taken longer.
"We would probably have had to disassemble and reassemble the entire PABX, which would have taken up to one week," says Templeton.
The added resilience of IP telephony comes from its inherently distributed topology, as Brunel has experienced at first hand. As core processing and switching functions of the analogue PABX are dispersed onto the call manager server, the data network and the IP handsets themselves, so resilience increases.
"There are multiple servers and multiple locations with multiple lines coming in compared to a single point of failure," says Templeton. "People are always concerned about the phone network going down. But with IP telephony a lot has to happen before the network goes down because it is so distributed."
Mobility of extensions, however, is the most tangible evidence of flexibility and has come in useful for a number of other contingency scenarios that have cropped up on the campus. Some of Brunel's buildings lie next to the river Pinn and flooding occurs from time to time. "We have a disaster recovery room and people are starting to use it more as they realise they can just arrive and plug their phones in," says Furber.
The burst valve proved to be an early test of the flexibility of VoIP. It was the first major disaster recovery scenario, and Furber hopes, the last. And it passed with flying colours. "We moved 300 people in one day," says Furber.
VoIP had originally appealed to the university as it sought to rationalise its networks and prepare for a digital future. In 2000, the IT department started a project to re-cable its campus with fibre, and whittle a medley of kit down to one supplier - Cisco. At that time it maintained Siemens' ISDX PABXs on the main campus and gradually introduced IP on to smaller satellite sites. The two worlds connected at a basic level to provide an integrated dialling plan.
The IT department also saw the upgrade as an opportunity to get an early taste of IP telephony and started trialling VoIP from Cisco in 2001. However, late delivery of Call Manager Version 3 software by Cisco meant a more comprehensive pilot of sophisticated routing and management options was put on hold. The university stuck to using the network as transport for voice traffic.
As a means of linking phone calls and of introducing new telephone extensions to the network in a flexible manner, IP telephony nonetheless proved to be a good investment. "The cost advantage was primarily being able to service new users without continuing investment in a system we had accepted we would replace," says Furber.
The pilot grew organically to 400 extensions. "If we ran out of cable on a copper bundle, or a new building was opened, we would go in with the data network and drop in an IP phone," says Furber.
In 2003, with more mature call manager software available, the university decided to move to one integrated voice and data network and to use the infrastructure in place to implement VoIP.
Furber and his team hired IBM for the procurement and consultancy and the installation. It picked Cisco because of its existing track record as a technology partner and the major piece of the VoIP upgrade was to move from layer two to layer three design.
The later version distributes functionality more widely, and so local boxes on the network play a bigger part in routing calls.
Read article: The big bang approach to VoIP migration