Distributed computing has rapidly been gaining popularity, with projects ranging from cancer research to the immensely popular SETI@Home project. However, with ever-increasing personal computer power and the growing availability and prevalence of high-speed Internet connections like broadband, it’s only natural that projects are being planned in almost every area of quantitative research. These fields include chemistry, molecular biology, epidemiology, population dynamics, climate modeling, and so on. Distributed computing has the power to revolutionize research by speeding up and reducing costs for data processing and analysis, and is already being used or about to be used in a wide variety of specific roles including:
• the search for prime numbers
• finding anti-cancer drugs
• cracking encryption keys
• finding optimal nuclear waste storage solutions
• finding cures for the flu
• reducing the side-effects associated with chemotherapy
• analyzing stock markets the world over
• a wide range of bioinformatics problems, including those related to the human genome project
Recent reports say that there are now over a billion Internet-connected computers, with an average of about 500 MHz of computing power each, yet SETI@Home, by far the most popular distributed computing project, has only about 500,000 active users. Clearly there is a tremendous potential for future growth and use of this new method of supercomputing.
One hot new area of computing is called grid computing, which is the idea that, in future, individuals, and especially companies, will simply plug into a network of computing power in a similar fashion to how they now plug into a grid of electric power. Thus, they would only need to pay for what they use, and not over-invest in expense hardware. Networks like this were recently announced in the United States and Great Britain. IBM is helping to build the British network, and The National Science Foundation, Intel, IBM, and Qwest recently contracted to build a $53 million network to be completed in 2003. It is expected to be able to do 11.6 teraflops (trillions of floating point operations per second) and hold 450 trillion bytes of data on an optical network that runs at 40-billion-bits-per-second. This could be used for more than just scientific research, but also commercial work that would be very cost effective. Some companies, such as Sun Microsystems (which has 3,000 such networked CPUs), already have a sort of internal network of distributed computers, called a computer “farm.” By distributing available CPU resources, they’re able to raise their percentage of CPU usage from 5-10% of the time all the way up to 80% or 90%.
Sun also recently released Grid Engine software, that greatly simplifies the search for idle CPUs. This technology, which makes computing-intensive tasks much more easy to accomplish will almost certainly become commonplace in businesses in the near future, many experts say. John Patrick, IBM’s vice-president responsible for Internet strategies, even went so far as to say, “I'm very confident now that the next big thing [for the Internet] will be grid computing” (Popovich).
As idle CPU time becomes more and more sought after, some companies, such as Distributed Science, say they are planning to pay participants to allow them to run distributed applications on the clients’ machines. In fact, it will allegedly be enough to cover basic monthly Internet connection fees. As of August 2000, the Distributed Science’s Process Tree Network claimed to have over 100,000 members. The project is currently to study the best way to dispose of nuclear waste, based on models of gamma radiation leaks from different storage containers. In addition to Distributed Science, several other companies, including Parabon and Popular Power have developed a slightly different distributed computing model from SETI@Home. Instead of having the client machines constant report results back to a host server, they share the results and pass the information from one computer to another in the network, in an innovative attempt to speed the process.
The potential for this be used for personal entertainment and social reasons, and also for commerical entertainment endeavors is quite apparent. First of all, distributed computing projects have the potential to be very funny, such as SIC@Home (Search for incredible coincidence at home), and YETI@Home (the search for bigfoot at home). The projects can also be people’s way of expressing their values and giving back to the Internet community, without even having to worry about technical knowledge. Having a screensaver that demonstrates the search for alien life can generate many conversations about our existence in the universe, and the merits of looking for proof that we are not alone, as opposed to using our CPU for more terrestrial issues such as looking for a cure for cancer. Beyond these personal issues, entertainment companies may soon explore distributed computing as a means of reducing costs. Pixar, for example, used a network of 100 computers for over 20 hours just to create certain intensive action sequences in Toy Story. Spreading that out even more over a network, using Grid Computing, for example, may be something they use to generate new films even more quickly and cheaply.
This is Microsoft’s big plan for our future. They hope to create a centralized set of software services that live on the Internet. We would need a subscription to access these services (Lucrative monthly fees AND no potential for piracy? It’s a dream come true!), and then whenever we needed Word, or Powerpoint, or anything else, we would just access it online. This would likely require such users to have high speed Internet access, but Microsoft said it would work on operating systems other than Windows, but didn’t specify which ones. Microsoft says it will use XML and Simple Object Access Protocol in order to deliver the applications. One key concept that .Net will introduce is the “universal canvas,” which will allow seamless application integration; for example, you could have an Excel spreadsheet in an e-mail message. The .Net concept is supposed to be ready between 2002-2004, and though not strictly distributed computing, it operates along very similar methodologies and will presumably manage the client demands by spreading out server demands on a distributed model. The model of networked, distributed applications has the potential to change all our notions about software and computing.
|[Home] [What Is It?] [History] [Future] [Concerns] [Efficiency] [Curr. Projects] [Resources]|