Grid computing is increasingly moving out of the research and academic environments that first adopted the technology, and becoming an important consideration for corporate IT projects.

One of the highest-profile examples of grids being used in a commercial situation is the infrastructure for Comic Relief's Red Nose Day online fund-raising efforts, a system which this year was also called into action to support the Live 8 anti-poverty concerts.

Warren Kerrigan, chief technology officer of technology design company Can Factory, was involved in the design, testing and operation of the system. He says grid computing was essential for dealing with the large but unpredictable peaks of traffic caused by two such publicly visible events.

'One of the things with a TV-based operation such as Comic Relief is that it's a very emotionally led audience,' he says. 'Without warning, a presenter will say: "I'm going to stand here for the next 20 minutes until we raise £1m," and people actually get up from their sofa and go to the web site, so you have some unusual traffic levels to deal with.

'When the BBC news comes on at 10pm, everybody gets up again and goes to donate, which is quite an interesting half an hour when a few million people try to give money online at the same time.' 

The web site developed for the charity runs on Oracle 10g grid-enabled database and application server software, using Sun Microsystems hardware, replicated at two separate locations for resilience.

On Red Nose Day this year, the site received more than 165 million hits from 1.3 million unique users, viewing 55 million page impressions, with a 10-minute average session time. It processed more than £4.5m of donations that were authorised online on the night, plus £1.5m more from interactive TV and a further £1.5m through call centres.

Extensive testing and building-in of various fall-back options for the web site were a vital part of the design.

'We only had one chance to get this right. If you look at a company such as Interflora, they have a Valentine's Day peak but if they go down during that week, they have the rest of that week to recover. If people miss the show on Red Nose Day they tend not to come back, so we only have one shot to capture this information, and it has to work every time,' he says.

'We also have a lot of preconfigured fallback points. As a user logs into the platform for the first time, we validate their address or pre-populate the field, check their credit card, and provide a video clip at the end to say thank you. As we start to get busier and run out of capacity, we stop validating the address or we don't show them the video clip. Also, as we go outside the peaks we immediately scale back up.'

And if there are any problems, the unique nature of the Red Nose Day television show is very useful.

'The people on the web site are also watching the TV show. We had a major disaster where we had to reconfigure the front-end switches and reboot some servers, so we got the presenter, Davina McCall, to go on national TV and tell everyone to stop going to the web site, and the traffic dropped straight down,' says Kerrigan.

'We made some very quick changes and then she went back on and said: "Thanks very much, everyone back on," and the site came back up, and it saved the show.'

Processing donations in real time on a high-performance grid infrastructure has additional benefits for the charity, says Kerrigan.

'All the donations are processed for free, but the bank was previously inputting them itself, so it could take about six months to receive all the money. If you imagine the interest on £32m over a six-month period, it is staggering. That's why this had to go into a different system and an automated process,' he says.

'We also have the problem of reporting back how many of those are false or not authorised, so the TV show can give out an accurate update on the pledges. We had security problems where people used stolen cards, so there were a lot of reconciliations to do during the show. We have a massive integrated grid platform that does live authorisation, which processed about a quarter of a million donations in six hours.'

This year, having achieved record online donations for Red Nose Day, the team supporting the infrastructure then faced an unexpected new challenge.

'Comic Relief organiser Richard Curtis and Bob Geldof were chatting, and Geldof said to Curtis: "We have this Live 8 thing; I don't suppose we can host it on that big platform you have?" "Sure, no problem," replied Curtis. Suddenly we had four weeks to change the way our system works. Instead of taking a quarter of a million donations in six hours, we had to take half a million petition signings and allow people to upload photos and a whole load of different things - with four weeks' notice,' says Kerrigan.

The flexibility of the grid infrastructure was vital to getting a new web site up and running for the day of the global Live 8 concerts in July. Kerrigan says testing was critical, although it had to be done as quickly as possible.

And with the worldwide publicity for the events, security was a major consideration.

'One of the challenges we had with Live 8 is that we had 250 million names and validated email addresses in the database,' says Kerrigan.

'That's quite a useful database for the spammer network to get their hands on. Equally, there are a quarter of a million credit cards on the database. So we had some interesting challenges.

'We tend to be a little clever: we often program in error messages from other operating systems or other web servers to give the impression that users are actually logging into another type of software. That helps to confuse the hackers.'

Warren Kerrigan was speaking during a live Computing Web Seminar on grid technology. The full seminar, including Kerrigan's presentation on the Comic Relief grid computing system, is available to view free of charge at: www.computing.co.uk/webseminars .