Tuesday, January 25, 2011

Choosing projects

I am attached to 10 BOINC projects and often ponder what resource share to allocate to each one to get a good daily credit as well as contribute some meaningful data to science.
While there are some tables around that list the credit/hour ratings of the various projects, that real-world results are probably fairly different due to a mix of operating systems, GPU vs CPU etc.

To get some indication of how to run my "farm" for 2011, I have decided to run all the projects with an equal resource share for a week. I have allowed each project to run its own applications (no optimized apps) and its choice of sub-project.

The projects involved are:

AQUA@Home
QMC@Home
PrimeGrid
LHC@Home
DNETC@Home
Collatz Conjecture
Milkyway@Home
Rosetta@Home
Einstein@Home
SETI@Home

The only project with modified preferences is Collatz Conjetcure. Its CPU application would take around 1200 hours on my slower CPU's, thus rendering a 1 week trial pointless. Collatz's ATI/Linux app also slows my PC down to a point where it is unusable, so that is excluded too.
As AQUA@Home is currently down and tomorrow is a public holiday in Australia, I have decided to start the evaluation on 31/01/2011. This will also allow the PCs time to run out excessive work units from certain projects that had high resource share.

No comments:

Post a Comment