Basically it depends on how you set up your virtual images. If you set each of them to use the multiple cores, then boinc will see them as multiple cores. Which would be redundant and cause them to fight for resources.

We do not recommend running it on virtual machines. The main reason for that is, BOINC runs at lowest priority on the system. The system in this case being the virtual machine. BUT on the actual machine the virtual machine is running at normal priority. This would cause it to use up all of the cpu. If you wanted to use the cpu for anything else it would have to compete with the vm images.

It is best to install BOINC on the base system that hosting the vm images. This would allow it to back off as lowest priority when the virtual machines required CPU.

Hope this helps,
-Uplinger

Well, but like i said, you would then have 4 processes using up 4 cores each running at 100% cpu without knowledge of the other one so fighting and using up 4 times the memory on the computer. The system would in theory become unresponsive.

Also, I would imagine that there would be a downside to the CPU usage in comparison to your benchmarks...

-Uplinger

"16 days of credit" meaning "16 days of Runtime credit", not points. With a Q6600 obviously there are X cycles to dole out, so whether you have them efficiently used by 4 science processes or via VM to run 16 processes is up front rather inefficient.

Whether you get 16 CPU days? Probably less as wallclock is not same as CPU time, but then on a hyperthreaded I7 you get ~8 days for 4 cores. How true the benchmarks will then be?

Yes it's been done but cannot remember what the credit like outcome was i.e. low benchmarks or constant outlier?

Hello Jean,
I think the basic question is : How well designed is Boinc ? ...
I can imagine that using multiple VMs seems to be "attractive" regarding an artificial increasing of crunching days. But, in this case, how will be credit estimated and granted.
I am afraid that such approach could have a significant impact of the "credit granting mechanism" and finally having a "negative" impact on the honest co-crunchers.
Cheers,
Yves

Virtual machines are designed (generally) so that the software running above it does not know that it is running in a virtual image. The operating system (usually) does not know and certainly applications running on top of the operating system do not. Thus BOINC could not know if it is running on a virtual machine or not.

However, in this situation the actual performance of the virtual machines would be poor and this would be reflected in the credit granted for the results returned.

If we find that that this becomes an issue then we could change the requirements for the medals to require a min cpu time and min credit earned. I wouldn't want to do this now because it would likely cause some people to lose medals, but this would address the issue.

How about this. Don't cheat. Don't make us waste our time fighting cheating when we could be working on other items. Focus on contributing to research. If you see someone else cheating - call them on it. Public pressure is often more effective then any mechanisms we could put in place.

thanks,
Kevin