|
8 July 2001 WORK IN PROGRESS!
ith the advent of networking and the Internet it is now possible to distribute processing over a network. An extreme example of this involves The National Foundation for Cancer Research (NFCR) Centre for Drug Discovery in the Department of Chemistry at the University of Oxford, England They are working on a project which uses CPU time of users on the Internet, when a registered computer uses their screen saver (http://www.ud.com). The program that runs on the computer is one which searches for new drugs in the treatment of cancer. By July 2001 this program had achieved 213,568,434 hours of CPU time using over 760,000 devices. This has achieved one of the world's largest computers. Most people understand the concept of not using up available local memory, or disk resources, or even network capacity, but one of the largest underutilized resources is CPU time. Many computers, especially in office, lie idle for many hours in the day. These computers could easily be performing other tasks, such as providing solutions to state-of-the-art research projects.
A good example of a centralized system against a distributed systems is in the banking industry. Figure 1 shows how a bank might want to organize its business. In this case it devolves decision making, account management and logistics to regional offices, which then devolves these to local offices. This allows for a distribution of the activities, and, for example, a holiday in one regional area will not affect the rest of the business. If the bank had a centralized model, all the customers, staff and logistics would be centralized in a single place. This would obviously be inefficient and would cause a great deal of strain on the central site. It would also not be possible to setup an efficient system so that every customer would be able to withdraw cash from the central site. A more efficient model is to use ATM's distributed to local offices. Most governments around the world operate with this distrib-uted model, where the central government create the rules and policies, which are then distributed on a regional basis with regional councils. These are then passed onto local councils, which implement the policies. A centralized government would create all the policies, then decide how these would be implemented, and then would be in total control of its implementation.
Distributed processing has many advantages over localized processing, especially in:
|

|
Using specialized resources, which would not normally be accessible from a local com-puter, such as enhanced processing or increased amount of memory storage.
|
|

|
Using parallel processing, where a computing problem is split into a number of parallel task, which are distributed over the network.
|
|

|
Reducing the loading on the local computer, as tasks can be processed on remote computers.
|

Figure 1: Centralized v. distributed (© billatnapier)
Chapter 6, Mastering Computing, W.Buchanan, Palgrave.
Comments on this essay
If you've got any comments on this essay (no matter if you agree or
disagree with it), please send them to me using the form below.
Note your comments may be published to a comments pages, but none
of your details will be added to the comment, just the date that it
was sent. Thank you.
|