When the University of Birmingham upgraded its core network a few years ago, it set in course a train of events that would lead to it becoming one of the UK's largest - and yet secure - campus wireless LANs.
The university now provides its 30,000 students and 6000 staff with both wired and wireless network access. It also provides wireless access to visitors and guests, and by the end of this year expects to have complete coverage over three sites, from offices and laboratories, through student bedrooms, and right across the playing fields.
The process began back in 2003, when the university put out a public tender for a new core network, says head of networks John Turnbull.
Its existing backbone used 3Com kit on an FDDI ring, and had grown to the point where the inter-switch links between the four core switches each comprised eight aggregated Gigabit connections. The IT services group realised that upgrading to 10G trunks would not only improve performance, but would also free up cross-campus fibres, some of them 2km or 2.5km long, and release a lot of LX GBICs for re-use.
"We had to go to EU procurement because the project was £6 million in tin, and another £6 million in fibre," Turnbull says. "The usual candidates bid for it, we produced a shortlist and asked for demo kit to put through serious testing."
Three companies obliged - Foundry, Extreme and Cisco, he adds. "Foundry had just released its Jetcore architecture [which became the FastIron range] and came out on top in performance, closely followed by Extreme. On ease of management it was the same, with Cisco way behind because it had three different operating systems."
But where evaluation turned into revelation was when the university IT team discovered flow monitoring: "We took a [Foundry] 4802 stackable and plugged it into our main network, and all of a sudden we could see what was going on - it was like we'd been blind before," Turnbull recalls. "We saw a lot of viruses, student downloads, and so on."
So one of the things that sealed the decision to build the new core around Foundry's FastIron switches was that Foundry had sFlow monitoring every port, yet with no discernable performance impact on the switch, while at that point Cisco had not yet built NetFlow into its ASICs, so flow monitoring on the Cisco switches produced a performance hit.