It is a question as old as the IT industry itself - what is the most cost-effective way of delivering IT to end-users? There are basically two choices: do the processing on the desktop or do it somewhere else. How do these alternatives compare? Dundee City Council did the sums and voted for centralising the processing while rolling out thin-clients on the desktop.
There was one overwhelming reason for the council's decision to adopt thin-client computing. "The huge cost is not the capital cost," says Ged Bell, IT engineering manager at Dundee City Council. This is instead reserved for that familiar, but fearsome, factor - support.
Desktop support costs were the main issue. "The primary driver was the need to look at the increasing problem of supporting disparate PCs, each with a slightly different configuration," he says.
With 4,500 PCs to support across all the departments, including social work, education, housing and planning, that was a considerable task.
"If we were a private company then I could pass on the charge of support, but it's not acceptable in a local authority to keep saying I need another three support staff, and then another three, and so on," says Bell.
So, in mid-1997, a deep breath was taken and a fundamental rethink of the options available for delivering applications to end-users was embarked upon by the council. It also coincided with the need for reorganisation and rationalisation of the IT infrastructure following the previous year's local authority shake-up, which saw the city council formed out of the old district council and parts of Tayside Regional Council.
"We took a fresh look at how we could minimise the cost of support," says Bell. "We took a lot of time looking at the market."
But Bell was well aware that "thinking thin" could make finding a ready-made solution harder. The issue of thin client versus traditional fat-client PCs is, he emphasises, "Very much a matter of horses for courses. Thin clients are not for everyone."
A critical factor in making the choice between the two options is just what end-users need to do with their computing power. The council looked at thin-client technology, and divided users roughly into two camps: those for whom thin-client computing could deliver as good a quality of IT service, but with the significantly lower support costs that were obtainable; and those for whom traditional PCs would be the best delivery mechanism.
The first category comprised most of the council's office worker end-users, who wanted access to a mix of general office software such as word-processing plus their specialist departmental software, as well as business studies students in schools. It also, ironically, included some areas of the old district council that had not been through the PC revolution, but were still working on terminal-based Office Power from ICL. They were the easiest to shift to Windows-based terminals.
The second category comprised technical staff such as architects, who needed high-level graphics processing at the client, and those needing high-availability system access, such as departmental directors' secretaries. It also included computer studies students who needed PCs in order to study things like operating systems.
Bell also agreed that PCs could be retained by one further type of user - the IT enthusiast found in every office to whom all the other staff turn for ad-hoc helpdesking, and who regard themselves as power users and like to configure their own machines. This was largely a compromise decision made in the light of real-world practicalities, as when the move to thin-client architecture was first mooted, there was a pro-PC backlash from this group of end-users.
"We met with significant resistance from the in-house IT gurus - not the people who work with IT, but the magazine readers - who looked for reasons for PCs not to be taken away from them," says Bell.
Getting buy-in from the end-user community meant respecting individual opinions and departmental politics. "We wanted to minimise the number of people with fat PCs, but there was obviously going to be some in-house politics to consider," says Bell.
He even, fleetingly, considered cheating. "At one point, we thought of putting Pentium II 333MHz stickers on all the Windows-based terminals just to keep users quiet," he recalls.
Instead, Bell took to the road to sell the new architecture, address end-user concerns and respect sensibilities. One issue he had to take into account, illogical though it may be, is that many workers come to regard their desktop PC as their own personal property.
"We had to remind them that it was a device on a desk, and a tool for doing their jobs - which was to serve the public," says Bell.
Overall, Bell and his team put a considerable amount of effort into winning over hearts and minds. "We spent a lot of time on it. We talked to departmental chief officers and their managers, and below that level too. We almost did a roadshow," he says.
The roadshow was variably targeted, according to audience. "The message changed depending on who I was talking to," says Bell. "To managers I talked pounds and pence, to techies I talked technology."
Although the amount of time and effort put into getting buy-in from users was questioned, Bell considers it not only worthwhile, but essential. "We couldn't just issue diktats about thin clients. We had to talk about the benefits and get them to realise what these were. By and large I'm very confident that overall the implementation has been looked on favourably," he says.
Part of the deal over getting buy-in has involved compromise on both sides. While Bell has respected that there will be some end-users who have valid reasons for keeping PCs, such as architects and power-users, he has also ensured that the PCs can all act as terminal emulators as well.
So, for the classes of software that are held centrally on the servers, PC access is via terminal emulation. This model also allows a rolling change-over to thin-client architecture to be achieved more smoothly, as the servers can support a mixed population of terminals and PCs running terminal emulation, yet still retain the server-side advantages of central licensing, upgrading and maintenance.
The move to thin-client computing was undertaken via an initial six month pilot project among 35 administration staff in the education department. It was then rolled-out to selected user communities within all the council's departments.
Thin clients are also replacing PCs in schools for some classes of students. Computer labs in Dundee's 10 secondary schools have been kitted-out with 20 Windows-based terminals per school, each school supported by a single server running Microsoft Office via Citrix Winframe terminal serve software.
"In the past," says Bell, "Business studies students sat at 20 stand-alone machines, and before each class the teacher had to reset each to its base configuration, spending 15 or 20 minutes in total each time. Now everyone is sharing the same set-up, which only has to be changed once when required. This means more time for teaching and less time spent on set-up and administration."
One of the most successful implementations has been in the social services department, replacing a client-server PC-based client index - K2 - first implemented in l996 on 150 486 PCs, with subsequent enhancements to the software which put the PCs under pressure. Now, the client component is running on about half a dozen Winframe servers spread across the remote sites. Not only does the department now have server-to-server computing, but disparate software versions have been standardised, making updates much easier and so dramatically reducing costs.
When thin-client roll-out is complete at the end of the financial year in April, Dundee will have a terminal user population of approximately 650.
"There were about 4500 PCs, and there still are," says Bell. But he also points out that not only are some of these PCs operating as terminal emulators as well as PCs, but the total user population is itself growing, both in departments and in schools as part of the National Grid for Learning programme.
The council's calculations of the total cost of ownership for thin-client computing are convincing, believes Bell. Although the difference in capital costs between PCs and terminals tends to be insignificant, Dundee struck a good deal with its suppliers, and got the terminals for around £450 each from local Wyse reseller Barron McCann. But then capital cost differences were never envisaged as the means of cost-saving through adopting thin clients.
The council has also been fortunate when it comes to networking costs. The inescapable corollary to thin clients is not just fat servers, but fat networks as well. But Dundee has managed to avoid having to assign the cost of network upgrades to the thin client programme. To begin with, points out Bell, network demand is kept as low as possible by situating servers as physically close to their end-user offices as possible. Thankfully, the two council office sites, in Dundee's original City Chambers and the tower block behind, are physically close enough for a fibre link to have been laid. Remote offices and sites take advantage of Dundee's cable infrastructure.
In addition, there is an ongoing cycle of network upgrade going on through the council as ISDN lines are gradually replaced by 2Mbit cable pipes. So far as TCO calculations are concerned, however, network upgrade costs are not included in the total cost of thin-client computing.
Comparing software licensing costs between terminals and servers versus PCs is complex, but Bell is confident that overall licensing is cheaper with thin-client architecture. In addition, the previous lack of standardisation in software, with different versions of e-mail and packages was imposing a significant burden on support costs.
When the final calculations were done, the TCO results were impressive and very convincing - annual support costs are down 35%.
This is a language end-users love to hear. Dundee council's IT department operates on a service-level agreement basis with end-user departments, who pay for their support. Now, thanks to the introduction of widespread thin-client computing, the 35% reduction in support costs has meant that their service bills, if not down overall, are up by less than they would have been otherwise.
In local authorities, more than anywhere else except the health service, less money spent on IT for the same amount of IT service, means more money available for the really useful things in a community. And this can only be a good thing.
At a glance
The organisation: Dundee City Council, formed in 1996 to include part of the old District Council and parts of Tayside Regional Council
The challenge: to meet stringent public sector cost and performance targets with existing PC-centric computer infrastructures
The solution: abandon PCs for a computing environment based on Windows NT servers and thin-client technology
Is there a downside to thin-client computing?
"The only thing about thin clients is that all your eggs are in one basket," points out Bell. "It means you are more dependent on the servers and on the network. We mitigate that vulnerability by ensuring that key users, such as directors' secretaries, have retained their PCs."
That means that although the standard log-in is to run off the server, PC end-users can revert to local desktop processing if necessary.
"But we very much try to discourage them doing that unless it's necessary," says Bell.
The bottom line for thin clients? Don't deploy them indiscriminately. "It's not a blanket solution," says Bell. But where they do make sense, "They result in sustainable support."
What the BuyIT experts say...
Chairman of IT faculty, Institute of Chartered Accountants in England & Wales
Cost saving is a strong incentive for change and can sometimes lead to imaginative and innovative solutions.
One of the cost benefits of adopting thin-client architecture is the opportunity to redeploy ageing computer terminals. With the life of an Intel-based computer down to two or three years at most, it is worth pointing out that a thin-client approach means that only the server will need upgrading, extending the useful life of a terminal to as much as 10 years and leveraging investment in existing computers.
Nevertheless, projects do not necessarily have to generate large-scale financial benefits in order to be effective. In the Dundee example, there are lots of small gains in productivity achieved that, when taken together, constitute a substantial benefit.
The timing of costs and benefits is crucial. In some cases, as for Dundee, benefits can continue to flow well beyond the cost period of the project. A post-audit of benefits carried out on a regular basis can favourably influence management's view of future IT projects.
Implementing an untypical application inevitably means greater uncertainty about actual costs and a real risk that benefits will turn out to be less than anticipated. Extra care should therefore be taken in researching and analysing the solution, as was done in Dundee.
Dundee council have taken a considered risk with relatively new technology (in 1996) and have reaped the operational and financial rewards. Good suppliers work hard to help their customers reduce cost of ownership and thin client is one route. Cost containment is vitally important in the face of constantly increasing user demand for better and faster access to systems and information. The Government's modernisation agenda will increase this pressure in the public sector.
There are important lessons we can draw from the Dundee experience:
Chairman, BuyIT Best Practice Group
Dundee council's move to thin-client technology was a brave one in the face of the industry's general unquestioning acceptance of conventional PC-centric architecture.
The majority of users in most companies only need access to a few generic and company-specific applications, and to e-mail. Thin client architecture enables provision of these basic facilities without the ever-rising cost of ownership and management that applies to client/server computing.
In the BuyIT guidelines (guideline 3: evaluating the options for sourcing the solution,) we recommend assessing all potential options against an informed understanding of available solutions. It may be that this solution does not suit all applications but it is certainly worth including in initial thinking.
This case study happened to focus on a public sector organisation, but it is relevant to any organisation that is facing the need to be more responsive to customer requirements with ever-tighter budgets - that's all of us, isn't it?
I liked the focus on achieving clear business benefits - thin-client technology has meant measurable improvements in performance, manageability and cost-effectiveness for Dundee council. That's good practice.
But, in the end, the success of this project lay in Ged Bell's extensive user consultation and focus on "selling" the benefits. As always, fully involving the users is the only way to win.