Raistmer wrote:
> 
>  >Do we "reward" for the result returned, or for the resource that a
>  >project has usefully utilised?

> Reward should be for result returned of course, not for resource used.
> 1) Resource usage  just can be faked by too eager for credits 
> participant. No "cheating" should be ensured.
> 2) What about case of "broken resource" ? I personally had such 
> "resource" few days - excessively OCed GPU.

I agree. However...

My concern is that we can't measure or put a 'value' on the returned work.

But also, we can't just blindly 'reward' for just the time a WU spends 
on whatever hardware, regardless of whether good or broken.

Hence, we should reward exactly just for whatever resource is /utilised/ 
to give a correct WU result. (As counted by flops, whatever-ops, 
bandwidth usage, whatever was needed just for that result.)


Note that this should give motivation to improve the Boinc scheduler to 
balance tasks so that the best (or just a good mix) of parallel resource 
utilisation is achieved so that a host maximises it's WU throughput and 
RAC. ...But that is for another thread!

Regards,
Martin

-- 
--------------------
Martin Lomas
m_boincdev ml1 co uk.ddSPAM.dd
--------------------
_______________________________________________
boinc_dev mailing list
[email protected]
http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Reply via email to