New and Improved [email protected] will Form the Backbone of Distributed Computing Network
BOINC - the "Berkeley Open Infrastructure for Network Computing" - is moving through its development phases, and a new version of [email protected] is being tested right along with it. BOINC, as some of you may recall, is the system being developed by [email protected] project director David Anderson and his team to spread the credo of distributed computing to fields beyond SETI. The remarkable success of [email protected], which quickly became the most powerful computing network ever assembled, made it clear that distributed computing could be used for many other computing-intensive scientific projects. BOINC will make it possible for researchers in areas as diverse as molecular biology, climatology, and astrophysics to tap into the enormous but under-utilized calculating power of personal computers world-wide.
In principle, scientists do not necessarily have to wait around for BOINC to be completed in order to make use of distributed computing in their research. They could launch their own distributed computing programs, and some indeed have done so. [email protected] and climateprediction.net are only two of the better known projects, dedicated to research in the fields of molecular biology and climatology respectively. But launching an independent distributed computing project is a complex and labor-intensive task even for professional computer scientists. For researchers in other fields it is a daunting undertaking, which would take precious time and resources away from the main focus their of research. They will, in most cases, avoid it. BOINC will change all that. With BOINC, the basic distributed computing infrastructure will be available to any scientific group that wishes to make use of this remarkable new technique. With relatively small changes, the basic BOINC format could be used to research anything from long term evolutionary changes to the search for gravity waves. Furthermore, BOINC will bring to these different projects an inestimable resource without which no distributed computing project can proceed - a large pool of PC users, willing to put their computers' calculating power in the service of science.
Currently BOINC is in its "Beta testing" phase, meaning that it is being tested by a limited number of users who are running the program on their computers. Within the next two months David Anderson and his team hope to expand this select group of volunteers from several hundred to around ten thousand.
For several months now, the Beta testers have been running a BOINC-based program known as "astropulse." This program searches the masses of data collected by the [email protected] receiver at Arecibo for brief but powerful electromagnetic bursts, signifying the collapse of black holes. In the past few weeks astropulse was joined on the volunteers' computers by an experimental BOINC-based version of [email protected] itself. In the long term, the current [email protected] platform will be phased out for all users and replaced by the new BOINC version.
What will the new [email protected] be like? In most respects it will not be very different from the current version, but it will allow for greater flexibility for both the [email protected] scientists and for the millions of users around the world who run the program on their computers. Most notably, the current [email protected] program is designed to analyze only data that fits the parameters of the equipment currenty used at Arecibo. For example, the program only looks for gaussians that last around 12 seconds, because that happens to be the time it takes the Arecibo beam to scan any given point in the sky. Similarly, it can only analyze data from a 2 bit recorder, because that happens to be the type of instrument currently used to record data at Arecibo, and so on. Any data that deviates from these strict parameters simply cannot be processed. As a result, the current [email protected] program can never be used to analyze data collected at any location other than Arecibo, or using instruments other than those currently in place.
This can be a problem. It could, for instance become a serious hurdle if and when [email protected] follows up on its plans to collect data at Australia's Parkes observatory, because the parameters and instruments on the Australian radio telescopes are very different from those at Arecibo. To analyze this data, [email protected] users would have to download a completely new version of the program, tailored specifically for the Parkes observatory. Once they did so, they could no longer process the old-style data originating at Arecibo.
Even now, when Arecibo is the sole source of [email protected] observations, the inflexibility of the program can cause problems. This became clear following the Stellar Countdown session at Arecibo last March, when in addition to [email protected]'s standard 2 bit recorder, Chief Scientist Dan Werthimer's team employed a highly sensitive 8 bit recorder as well. The analysis of the 2 bit recordings was completed quickly, by sending ordinary work units to users around the world. The 8 bit recordings, however, are still unanalyzed because they cannot be processed by the standard [email protected] program, installed on users computers.
This will soon change, says David Anderson. In the new BOINC-based [email protected] parameters such as a radio telescope's beam-width or a recorder's level of sensitivity will not be "hard wired" into the program. Instead, they will be part of the information provided with every work unit. A standard work unit from Arecibo will instruct the program that the data was recorded at a 2 bit sensitivity, and that the telescope's beam-width - and therefore a gaussian's duration - is 12 seconds. A work unit originating at Parkes, or recorded at a higher resolution, will carry with it different parameters and the [email protected] program will adjust itself accordingly.
[email protected] and BOINC are gradually converging, and the benefits for both are substantial. While [email protected] enjoys the increased flexibility of the BOINC platform, it brings to BOINC something of inestimable value to a distributed computing project: millions of [email protected] users, willing to use their computers' processing power for the advancement of scientific research.