As you read, watch TV or mow the lawn, your brain is hard at work. Beneath the surface of conscious thought, billions of neurons are silently networking, analysing information, formulating ideas and exploring problems.

Everyone can remember times when, as if from nowhere, a novel idea or a solution to a problem they had been thinking about earlier, suddenly popped into mind. The brain has simply been using its idle processing capacity - out of sight, out of mind.

Deakin University computing scientist Professor Andrzej Goscinski is developing a revolutionary operating system that will allow a cluster of computer workstations to work in a similar way, giving each workstation the performance of a supercomputer by harnessing the idle processing and storage capacity of other workstations connected by local and/or wide area networks.

It would all happen transparently, in the background - the operator would only notice that their computer was performing computationally intensive tasks much faster, or had acquired the power to solve problems beyond its normal processing capacity.

"Many computations are so intensive that they cannot be completed on a single workstation in a reasonable time," Professor Goscinski said.

"The system we are developing would provide a programmer with enormous computational power, as well as access to services that might not be available on their local workstations. It would make life much easier."

"Our system would search the network until it found enough idle workstations to perform your particular task, carry out the computation on these workstations, and return the results back to your computer.

"To do that with current technology is very difficult and time-consuming. The challenge is to make it fast and simple.

"We also need to provide reliability - people are driven mad if their computer crashes, so if we send tasks elsewhere in the network, we need to know we can retrieve them, along with the results, without losing the lot."

The system would be applicable not just to powerful workstations, but to personal computers. "There are enormous numbers of PCs sitting on desks, and they can be idle for over 70 per cent of the time, particularly late at night and early in the morning," Professor Goscinski said. They can be used as a very powerful supercomputer. For example, the networked computers in a retail shopping network could by night perform data mining - sifting through enormous volumes of sales data to identify patterns of consumer spending.

"The idea would be to manage the task in such a way that the application could be executed in parallel on all the idle workstations, to speed up processing for instance by a factor of 10 or more times when using 20 idle workstations."

Such a system would offer enormous cost savings for companies. The price of a massively parallel supercomputer begins around $1 million, whereas it would cost only one hundred thousand dollars to link 20 workstations using a local area network into a system with comparable computing power - and many companies already own the computers.

Professor Goscinski and his colleagues have been working on an operating system to support distributed parallel processing that can be easily used by any programmer, instead of requiring a network "guru".

With today's operating systems, says Professor Goscinski, the computer and network expert needs to specify the location and other technical details of remote machines very precisely.

"In the area of managing resources of distributed systems we haven't really advanced much beyond the 1970s, when only gurus could understand computers," he said. "Current operating systems were never designed for distributed and parallel processing.

"We want programmers to be unaware that their application is running in parallel on a computer cluster - we want to hide the operation completely, develop the software to exploit very fast communication networks, and provide tools that the operator has already become familiar with in ordinary sequential programming.

The Deakin team have been testing their system's performance and reliability by executing computationally intensive tasks on the department's own internal computer cluster.

They have discovered that any distributed system must inevitably trade off speed against ease of use - in this case, the need to provide the user with such simplicity and transparency that they are unaware their task is being run on a computer cluster.

Instead of having to specify precisely which machines are to be used, the Deakin researchers have already implemented a system that automatically identifies idle capacity on other computers, and brings them into the network.

"If someone comes in and begins to use one of the computers in the network, the system automatically moves the computation to another machine, without the user being aware that their machine was being used."

The system makes use not only of the processor and memory on the remote machine, but its hard disk. The remote computer automatically relays its results back to a server which returns it to the user in the form they specified when they initiated the task on their own machine.

"Even though the user has no knowledge of where the task was executed, and how it was done, everything remains available to be retrieved, transparently, whenever the user requires it.

"It's a very difficult challenge, but it could change computing enormously. There are major advantages in proceeding this way - we already know this from systems that do the same thing manually.

With the explosion in electronic commerce, many businesses are already doing their banking by telephone or the Internet. Professor Goscinski's research group is collaborating with Professor Webb and Garner's research group in a project that seeks to integrate electronic commerce, knowledge acquisition, management systems and global computing based on clusters of computers into a seamless whole.

To make it all work requires networks much faster than any available today - computers would need to exchange data at rates around a gigabit per second, compared with the current 100-megabits per second limit for a system like Fast Ethernet.

Professor Goscinski is already developing an improved version of his revolutionary RHODOS (ResearcH Oriented Distributed Operating System), which, when he and his team completed it in 1995, was recognised as the world's first fully mature distributed operating system.

His research team is at the international vanguard of research that would revolutionise computing by making transparent, distributed and parallel systems a reality.