Follow Us.
Become a Fan! 
A thin client system needs, most of all, on a reliable server. At first glance, this would seem to be the weak link in the entire installation - lose the server, and you've lost the whole system.
One of the classic arguments against moving to a thin client system (where multiple clients are run from a single NT or Windows 2000 server) is that if the server goes down, none of the attached clients are operable. However there are very few distributed systems where there is not some key PC somewhere that is essential to the systems operation.
The most common single point of failure in a traditional distributed system is the I/O server. Most plants that use any type of Fieldbus communication cannot afford the expense of having a Fieldbus I/O card in each of the control and monitoring PCs. A common approach is to use one PC to poll all of the devices, and have it serve the data to any applications that need it. In that case, a failure in that PC will result in a loss of data throughout the entire plant.
Other systems use a single PC to collect and retrieve data from a database. PCs throughout the plant make SQL calls to add records to this database, and then make more SQL calls to get data back out. Or perhaps a single PC is used to access the network. Either way, most distributed systems are more dependent on a single PC than they realize.
A thin client system can actually be more reliable than a traditional distributed system. This is done by making sure that the server is protected. There are several different methods of doing this, each of which has different costs and levels of protection. We will discuss each of these in turn.
The least expensive solution to server reliability is to use more than one server. Each server is set up to be able to support thin clients, and then different clients are attached to different servers. The servers are kept current with the same applications, either manually or automatically. If one server fails, the clients that were using that server can be moved to the other servers.
The next step in server reliability is to use special thin client Load Balancing software. Like the Server Group above, this approach uses multiple servers that contain the applications. The difference is that now, instead of clients being manually assigned to individual servers, the clients are attached to servers based on the current load of each server. When a client is powered up, the Load Balancing software looks for the server with the lightest resource load, and automatically assigns it to that server. To restore client operation after a server failure, simply reboot the clients - they will automatically be routed to a working server.
The best way to assure that there is no single point of failure in a thin client system is to use servers that have been designed to run continuously. With one of these servers, any part that fails is immediately switched to a working part on the other server, with the application never knowing the difference. Additionally, these servers bring redundancy to each client in the plant. To achieve this level of reliability in a traditional distributed system, each PC throughout the factory would have to be made redundant. Because of the nature of a thin client system, the same effect is achieved by spending the expense only once - on the server.
For more information on ACP Industrial Thin Client computers, please visit our web site at http://www.thinmanager.com
To sign up for the E-mail newsletter go here: ACP newsletter signup
For an archive of past newsletter articles go to: ACP Newsletter Archive
Copyright © Automation Control Products, Inc. All rights reserved.
