Information Age: News, analysis & insight for IT & business leaders

Most mid-sized and large companies rely on data centre services for the smooth running of every business day. The workloads placed on the data centre tend to show occasional peaks and troughs throughout the year; for the most part though, demand is reasonably steady and predictable. Almost the exact opposite is true for the charity Comic Relief.

Comic Relief holds a marathon television event, aired on the BBC, every two years. The demand on its data centre starts building a few weeks before the Red Nose Day telethon; on the evening itself demand spikes off the scale.

The Comic Relief data centre has to support a call centre, an interactive TV portal, and a web site. Its main purpose is to provide transaction systems that can process all the pledges made over the preceding weeks and especially on the night.

In 2005, web traffic was moderate in the weeks leading up to the telethon. On the night it got 530,000 hits.

"We have just six hours on one night to do it," says Gill. "The BBC aren't going to reschedule if the data centre gets a bit warm, or the kit falls over. It has to work."

This sort of business demand would challenge most data centre managers, but Martin Gill, Comic Relief's new media manager, has a further problem: he has no data centre budget. The charity makes a 'golden pound' promise whereby 100% of the money donated will reach good causes.

Like many good causes, Comic Relief can rely on the goodwill of technology vendors to support it, in this case Oracle, Sun Microsystems and design company Can Factory. But despite their assistance, it still faced a technical challenge to ensure it had robust, scalable systems. "We can't afford to take any risks," says Gill.

Comic Relief opted for an 'N+1' redundancy model, building two interconnected data centres, set 11 miles apart in Leeds. Each was filled with clusters of Sun 880 servers running Oracle's 10g application server and a transaction processing system built with partners WorldPay and Zeus Technology.

This model also supports the instantaneous scalability required to withstand the steep rise in traffic that Comic Relief's web site receives on the evening of Red Nose Day. Dynamic management is crucial in order to handle any unforeseen challenges, as is having a plan B. During the 2005 telethon event, one of the servers in the data centre failed - given the absolute requirement to provide an uninterrupted service, this had the potential to be disastrous.

However, by using dynamic load balancing, the IT team were able to move donor sessions across to the other site without any loss in data. For Gill and his team, every Red Nose Day is a trial by fire for the latest incarnation of the IT infrastructure. In 2005, despite the one hitch, Comic Relief was able to automatically process more than 242,000 online transactions (over a terabyte of data) in six hours and raise a total of over £7.5 million through its new media operation.

Gill believes that building the system from hardware and software from best of breed suppliers is the most successful approach. Having considered partnering with a utility computing provider who could manage and host the whole infrastructure, he believes that the ability of such providers to manage infrastructures of the size Comic Relief requires has not yet been proven. "Secondly, it's too much commitment to ask of just one company," says Gill.

Resource constraints mean that Comic Relief must use the same data centres to perform both pre- and post-event functions such as distributing on-line media and counting the money.

The 10g application servers intelligently distribute processing load across the data centre so as one task winds down, the same hardware can be reassigned to the next, letting Comic Relief's IT team 'reuse kit on the fly.'