HOW DISTRIBUTED COMPUTING WORKS

Let us say you want to build a house. If you hired a single worker to do all the work, no matter how good and fast the worker is, it will take a long time. However, if you distributed the work among many groups of workers: plumbers, roofers, electricians etc working independently of each other they can finish the work much faster. Distributed Computing is to distribute heavy computational load into thousands of computers that work when they are otherwise idle across the Internet.

Human Genome is the "blueprint" for proteins. But just like a blueprint is not enough to build an airplane, we need to understand better how proteins assemble themselves by folding. We do have computational methods to simulate protein folding. However, current computer speeds (even super computers) allow us only to simulate a nanosecond (10^-9) of this process, where the process of protein folding can take more than a microsecond (10^-6). Thus there is a thousand fold gap between the current computing speed and the computation power needed to simulate protein folding.

Folding@home has developed a new way to simulate protein folding which can break the microsecond barrier by dividing the work between multiple processors. By using a novel approach to create the Folding@home clients and server a near linear speed curve was achieved (for example: twice the number of computers twice the speed). As the number of processors increases so does the speed with which we fold proteins. Thus, with thousands of processors, we can break the microsecond barrier and unlock the mystery of how proteins fold.

Pande group breaks the total computations into small chunks called work units. Work units are assigned to computers that join folding@home. When the computer is idle, it does computations in the background. When the work unit is completed and internet connection is established the results are sent to a data base at Stanford servers. Researchers can access these data bases and extract valuable information. See also research paper summary.

Image courtesy of www.thinkquest.org

Right click (PC) Alt-Click (Mac) for the Flash Menu (Play, Rewind, Stop etc.)

Author: Tug Sezen