#1247 1999 · SETI@home (UC Berkeley Space Sciences Laboratory) · Astronomy / distributed computing
Berkeley couldn't afford a supercomputer for alien signals, so it borrowed idle home PCs
the problem
Searching radio telescope data for alien signals needed far more computing power than an astronomy budget could buy
background
The Arecibo Observatory's radio telescope collected roughly 35 gigabytes of raw data a day as part of the broader SETI (Search for Extraterrestrial Intelligence) effort, and searching that data for the narrow-band signal patterns that might indicate an artificial transmission required computing power well beyond what UC Berkeley's Space Sciences Laboratory could afford to buy or rent on any sustained basis. Traditional supercomputer time was expensive and had to be shared across many competing research projects, making a continuous, dedicated search impractical on the project's budget.
David Anderson and Dan Werthimer's team, developing the idea through the mid-1990s, noticed that millions of personal computers worldwide sat mostly idle, running screensavers that did no useful work while the owner was away. SETI@home, launched May 17, 1999, replaced that idle screensaver with software that downloaded a small chunk of Arecibo's radio data, processed it locally using the computer's spare cycles, and sent the analyzed results back automatically, donated for free by whoever installed it.
what everyone would do
The available options were applying for shared time on an institutional supercomputer, which was scarce and contested across many competing research projects, or trying to raise dedicated funding to purchase computing hardware the lab's astronomy budget couldn't otherwise justify.
what they saw
Anderson and Werthimer saw the world already held idle capacity, millions of PCs behind screensavers, and the missing piece was routing small chunks of data to that free capacity instead of buying more machines.
the move
SETI@home distributed small chunks of raw Arecibo radio telescope data to any volunteer's home computer running its free client software, which processed the data as a screensaver during the computer's otherwise-idle time and automatically returned the results, aggregating the donated spare computing power of hundreds of thousands of ordinary PCs into a single distributed search.
why it works
Because each unit of work was a small, independent chunk of radio data that could be processed and returned without any coordination between volunteers, the system could add new participants without added marginal management cost, scaling the total available computing power with how many people chose to install the client rather than with any fixed hardware budget. Framing participation as a passive screensaver rather than an active task lowered the cost of contributing to nearly zero, which is why adoption reached a million users within months of launch.
the payoff
Within three months of its May 1999 launch, a million volunteers had donated a combined 51,000 years of processor time.
where it breaks
The approach depends on a computational task that decomposes into small, independent units needing no real-time coordination between contributors, and on being able to verify returned results are correct without re-running the whole computation centrally; tasks requiring tightly coupled, sequential computation, or where a malicious or faulty volunteer machine could quietly corrupt results undetected, don't distribute well this way.
what came after
SETI@home became the model for volunteer distributed computing broadly, its underlying BOINC software framework later powering dozens of other scientific projects in fields from protein folding to climate modeling, before the SETI@home project itself paused sending new work units in March 2020 after 21 years of continuous volunteer computation.
references
- [1]Over One Million Served: SETI@Home Reaches A MilestoneScienceDaily, 1999sciencedaily.com