For several reasons, many offices and individuals leave their computers on all the time. Nightly backups, remote access, minor server functions and the desire for “instant-on” access is only a short list of the arguments for a computer to be on 24/7.
Modern computers already spend almost all their time waiting for the user to press a key or move the mouse, only occasionally being asked to work at their capacity. Games and 3D rendering are two common applications which are actually CPU bound, but the hardest many office computers ever work is when they’re rendering a 3D screen saver.
To take advantage of all this “wasted” processing power, several groups have had varying degrees of success in producing distributed computing projects. Client software is installed on a machine with cycles to spare, and it coordinates with a central server on the Internet to get a small piece of a very large problem to work on, known as a work-unit. The software then spends a few hours, days or in some cases months working on its assigned work-unit, reporting back once it’s complete.
One of the oldest projects is The Great Internet Mersenne Prime Search (GIMPS), started in 1996.01. The project looks for primes of the form 2^n - 1, and have found the four largest known primes. Information about the project and client software is available at (URL:http://www.mersenne.org/).
A project which is quite well known among “geeks” is Distributed.net. Started in 1997.03, the first effort for Distributed was the RC5-56 crypto challenge, a $10,000 prize offered by RSA Data Security for the first person to break a specific message encoded with their 56 bit public-key encryption. They broke it in only 212 days.
Right now they’re working on RC5-64, plus are looking for Optimal Golomb Rulers. They currently have the computing power roughly equal to 46,000 Pentium III machines clocked at 1000Hz running all day, every day. Details and clients for your computer available at (URL:http://www.distributed.net/).
For those who prefer a little eye candy with their donation of cycles, the SETI@home project is a good match. SETI stands for Search for ExtraTerrestrial Intelligence, and SETI@home is a distributed project to help process radio-telescope data searching for interesting signals. SETI@home works on such a large dataset that it only operates as a screen-saver, where-in it draws a very colourful display of what it’s currently working on. (URL:http://setiathome.ssl.berkeley.edu/) has everything you might want to know the effort and how to join.
What does this have to do with business? Well, the above projects, among others have been a bit of a proving ground for distributed computing enabled by the Internet, and corporations have started appearing offering a commercial dimension. For those who need large computational resources, such services may offer a cost effective alternative to buying or renting time on a super-computer. For those running the clients, it may generate a few dollars of revenue each month to help offset the electrical bill.
Popular Power is one such entity, located at (URL:http://www.popularpower.com/). Their first project is a non-profit effort to help battle the influenza virus. Another player is Parabon Computation, located at (URL:http://www.parabon.com/), who are currently working with the National Cancer Institute to analyze data on drug interaction with cancer cells.
The attraction to the CPU cycle “consumer” is clear. Rather than investing in quick-to-depreciate hardware which must be kept busy to pay for itself, cycles are purchased as needed; discounts for low-priority jobs. A small computer animation or engineering company, for example, could simply buy cycles as they need them. It’s interesting to note that the computing resources expended by some of these distributed projects on different efforts are greater to that needed to render the computer generated “Toy Story” movies by Pixar.
For the CPU cycle seller, or “provider”, hardly anything is noticed. When the computer isn’t dealing with the user, it’s busy figuring out how a protean folds for a drug company or drawing a very small section of a frame in an animation. Except in power-saving modes, CPU cycles can be thought of as a lost resource if they are not used doing something worthwhile.
For those computing problems which can be broken down into little chunks, distributed computing is likely to be an ever-more appealing solution. Moore’s law of doubling computational ability while the price drops in half every 18 months now appears possible to continue for another 30 years. There will be so much computing power available, distributing processing may be the only way to use it at all effectively.
Published in the Victoria Business Examiner.