Dat er wat discrepantie tussen de projecten onderling zit voor wat betreft de credits, mag bekend zijn.
Ook zal over het algemeen een Linux systeem wat meer punten krijgen dan hetzelfde systeem met Windows.
Dan is er nog het verschil in credits tussen 32 en 64 bits WU's indien een project die beide verschillende WU's aanbied voor de verschillende OS en hardware systemen.
Er zijn verbeteringen aan te brengen in het huidige credit systeem bij WCG, maar dat zal niet zomaar gebeuren zonder overtuigend bewijs.
Our position on this stems from fact that the problem of assigning credit is really difficult. Awarding credit is an attempt to assign a value to an unknown and varying amount of work processed by a untrusted computer that reports its own assessment of its computing power and its own measurement of how long the job took to run. This is a problem and challenge that has existed since beginning of volunteer computing. It has also been EXTENSIVELY debated and discussed.
The system that is in place now works well enough on average over time. This is not to say that it is anywhere close to perfect (it isn't), but it works well enough most of the time.
If there are people who have proposals to improve the credit system, they are encouraged to work with the BOINC project to make those changes. Note that many systems have been proposed over time and the overwhelming majority of them perform much worse then the current system. As a result, if you bring a proposed solution to them, you would need to be able to provide data and simulations of the performance of the proposed mechanism. The proposed solution would need to consider the range of BOINC projects as well as the existence of bad actors who try maximize points through ill means. It also needs to consider the fact that some projects use mostly integer based computations while others using floating point computations. Some run on GPUs and some on CPUs. Some are IO intensive and some use little IO. Some can run entirely in the CPU cache and some make extensive access to main memory. The proposed solution has to handle all of that complexity.
[
Voor 28% gewijzigd door
m_p_g op 02-09-2020 11:40
]