Volunteer computing, in which computer owners donate
processing and storage to scientific research projects,
creates massively parallel computers more powerful
than any supercomputer or cluster.
The BOINC project, based at the UC Berkeley Space Sciences Laboratory,
is developing middleware that makes it easy for
scientists to create volunteer computing projects,
and for computer owners to participate in these projects.
I will outline the design of BOINC and will focus on three areas:
1) why the BOINC scheduling server is so efficient;
2) why BOINC's credit system is important,
what its goals are, and how it works;
3) the potential capacity of volunteer computing,
in terms of FLOPS and bytes.