By continuing to use the site you agree to our Privacy & Cookies policy

Slimming the system

The language of technology sometimes defies translation, but the recently revived term 'thin client' turns out to be apt

Thin client? The world of computing has taken over a whole raft of architectural terminology from 'software architect' (a higher form of software engineer) to 'systems architect' and now 'thin clients'. Actually, thin-client architecture is a pretty old idea that is used on mainframes and many UNIX systems. But it is now becoming popular again at the non-mainframe level. This time it is driven by the relatively high cost of re-equipping fat client networks and the possibility of using free non-Microsoft administration software.

Thin-client is a way of describing a computer network in which the desktop kit used by people (the client) is minimal (thin). Most important computery things are done on a central computer, the server. So your staff members have screens, a simple computer to handle the network and input devices, such as mice and pens and scanners and digital camera images, and the odd output device, such as the screen and a personal printer or plotter.

The heavy-duty applications such as AutoCAD or MicroStation and Microsoft Office sit on the server's hard disk (or in bigger companies, on the hard disks of a server farm), where all the number-crunching and shading and parametrics and dimensioning take place - in short, all the processing activity for the whole office. Servers therefore need to be pretty powerful and need dedicated thin-client software such as Citrix's MetaFrame.

There is no particular economy in application software such as the above word-processing and CAD programmes, because you have to pay licence fees for them per seat in the usual way. And if a server falls over, everybody sits on their hands until it is brought back into service: the cheap individual desktop boxes can't operate as stand-alone computers.

On the other hand, in a more typical fat client network, where the desktop computers will have the primary everyday applications mounted on their hard drives, people still tend to have to sit on their hands when the servers fall over because all networks are quite interactive - even if it is only to the extent of running automatic back-ups and internet connections.

The savings in cost associated with thin-client networks come from not having to pay much for the desktop kit, which currently must be at its lowest price in the admittedly short history of computers. It is also to do with using Linux and, thus, almostfree server software - although a lot of network engineers still use Novell and especially Microsoft's XP Embedded.

The saving in desktop-kit costs occur because you are buying terminals for staff, rather than personal computers, although it could be difficult for the casual visitor to an office to spot the difference.

The big advantage for network administrators is that they can carry out upgrades and maintenance on the server rather than on every architect's computer. And that, potentially, saves an enormous amount of time.

When thin-client networks are wired up, security is said to be better because the serious stuff occurs securely inside the server. If someone wants to steal data they have to attack the server, which normally will be in a secure location. The most interesting facet of thin-client architecture is that it can be operated over wireless networks using both WiFi and 3G cellular networks. The possibility is now open for diskless - and therefore quite cheap - tablets, notebooks and handhelds to form part of the network. Maybe this means the end of security scares when people leave laptops in taxis.

But cheap is relative. The M1400 Centrino and Base, released in the US early this year by pioneering thinclient laptop manufacturer Motion, are actually around the price of ordinary tablets - but it is early days.

Wireless is slower than cable, so most CAD work will be run over wires, and, because of the processingintensive nature of CAD, the server has to be up to the job of processing tasks simultaneously for everybody on the network. It is also true that the thin client is inherently slower than working on a dedicated computer, so some advocates believe thin client is unsuitable for design, engineering and architectural practices.

But this is to play the numbers game rather than finding out how things work in real life - and there are primary factors such as the capacity and power of the server or servers, the number of people using the internal network, the intensity of standarduse patterns and the like which have to be factored in against a diminution of administrator time and re-equipping costs. But perhaps the greatest human problem for thin-clients is that people are inclined to resist the company IT nerds who seem to be taking over that private fiefdom that is their current personal computer.

Have your say

You must sign in to make a comment.

The searchable digital buildings archive with drawings from more than 1,500 projects

AJ newsletters