Author: Merlin SF, created 21 February 2008
BOINC, or Berkeley Open Infrastructure for Network Computing, is a remarkable phenomenon. It is the best-known name associated with distributed computing. The name also points to its origins at the University of California, Berkeley, one of the world’s leading centres of research from quantum mechanics to astrophysics.
Modern scientific observation produces an ever-growing amount of data for further analysis. One possible response is to construct a dedicated supercomputer. Imagine the hardware needed for the algorithms involved: in early 2008 a popular quad-core processor delivered about 5.5 GFLOPS, while a typical project could require tens of thousands of times more computing power. Ten thousand networked machines would require an enormous building and a multi-million-dollar investment.
Such a facility would also consume vast amounts of electricity and release large quantities of heat. Cooling, lighting, operation, maintenance and continued hardware upgrades add further costs. As the amount of scientific data grows, the centre would soon need more floors, more machines and another round of investment.
Distributed computing offers a different answer. Many people already own a computer that alternates between short periods of work and idle time. Even while switched on and consuming energy, its processor is often not fully occupied. Those unused intervals can process parts of scientific datasets.
Connecting many such computers creates a large, flexible computing structure. Small contributions of otherwise idle processor time add up to computing power that can surpass a single central machine.
BOINC divides large datasets into much smaller work units that ordinary computers can process over several hours or days. BOINC Manager and the BOINC core communicate with a selected project automatically. After joining, the project downloads its scientific application, algorithms and work units.
BOINC runs in the background and monitors the resources available to it. It computes when resources are free and yields when the operating system or an interactive application needs them. A completed work unit is uploaded and replaced by another one according to the participant’s settings.
Participation is voluntary and is not financially rewarded. Completed work produces virtual credit used in project and team statistics. Participants choose which projects to support and how many resources to provide. The reward is knowing that their computers contribute to research across many scientific fields.
Information about individual scientific projects is available in the Czech National Team project catalogue.