Locking the desktop in the filing cabinet

There are two main approaches to running PC applications in a corporate environment: desktop and thin client. In desktop systems, processing takes place on a machine local to the user. Data and applications may be stored centrally, but an autonomous PC placed a keyboard cable's length away from its user does the work. Thin-client systems have a dumb local box that shuttles keyboard, mouse and video signals between the user and a central server.

Desktop blades are a halfway house between the two approaches. Every user has a dedicated PC, kept and managed in a server centre, with a small box on the physical desk that connects to screen, keyboard and so on. The desktop blade PC has its own hard disk and operating system -- typically Windows XP, although in principle it can be any PC OS -- and is connected to the normal network resources such as file and print servers, Internet access and so on through normal networking.

The original blade desktops used special analogue techniques to send the video from the PCs up to a couple of hundred metres along twisted pairs to the user. This meant a minimal amount of modification to the PCs and the software they ran -- from an electrical and software viewpoint, they were just computers with very long leads -- but imposed significant physical restrictions on the scope of the installation. For some specialist applications, where security considerations limit the physical location of the userbase, this is no disadvantage, but it has limited the applicability of the idea in general. The latest generation of blade desktops uses more conventional remote access techniques akin to VNC or other remote control products: this puts input and output into IP packets which can then be switched and routed alongside any other network traffic.

The advantages of a desktop blade configuration are the security and manageability of a thin-client system, coupled with the individual performance guarantees and ability to run a wide variety of software that come with one PC per user. The user -- or anyone else without access to the data centre -- is prevented from walking away with the PC; it's also easy to set things up so that viruses or unauthorised software can't be introduced onto the network, because there is no local storage as standard. However, current blade desktops can be set up to provide USB connectivity to users, which needs careful management regardless of where the processor lives.

The disadvantages are a cost per seat greater than that of thin client -- where one server can be spread across multiple users -- and a reliance on one particular vendor's PC hardware, leading to lock-in and potential problems when upgrades are next due. Blade desktops do accept standard memory and disk upgrades, but lack standard PCI expansion slots.

American company ClearCube Technology was first to sell desktop blade systems. The company has installed desktop blade systems at financial institutions and at some military installations. IBM has resold some ClearCube systems, but HP is the first major manufacturer to develop and market its own blade-based desktop system as part of its Consolidated Client Infrastructure (CCI) initiative. So far, the blade concept has been mostly used to squeeze servers and communication equipment more efficiently into racks, but HP makes the case for extending the idea into desktop systems to help patch, licence and OS installation management, ease back-up and restore problems, and to maintain locality of sensitive data -- that is, to stop users making copies of secure data in insecure places.

Some of these ideas are reflected in lower management costs and, say desktop blade systems makers, in a lower total cost of ownership. However, it is hard to show significant TCO reduction in many instances, making desktop blades most attractive for the increase in reliability and decrease in downtime they promise. The idea is that if an individual blade does develop problems, it is faster and easier for a centralised support team to find and fix the issue if everything's in one place -- and as they appear to the network to be normal PCs, it is also easy to apply standard failover and management approaches to them. ClearCube is more aggressive about claims of downtime and cost reduction, saying that with its hotswap technology, a user will experience 99.9% availability and operating costs will be reduced by "at least 40%".

Desktop blades are unlikely to become a major feature on the enterprise hardware landscape in the near future, if only because of lock-in worries and the lack of flexibility they can bring. If the industry can standardise their physical attributes and manageability, so that blades from one maker can be swapped with those from another without trouble, then IT managers may feel they are sacrificing fewer future options and will be more comfortable with moving to this architecture. There will always be installations where the increased physical and electronic security and reliability of blade desktops will be recommendation enough: conversely, there are plenty of situations where the idea is not attractive. For the average IT manager, however, the TCO and manageability claims will need substantial proof before desktop blades become an attractive option.