boinc/doc/validation.html

52 lines
1.8 KiB
HTML

<h2>Accounting and result validation</h2>
<p>
Participants are given <b>credit</b> for the computations
performed by their hosts.
These credits are used to generate "leaderboards"
of individuals, teams, and categories (countries, CPU types, etc.).
It is to be expected that users will explore ways of "cheating",
i.e. getting undeserved credit.
<p>
In principles credit should reflect network transfers and
disk storage as well as computation.
However, it's hard to verify these activities,
so they aren't included in credit.
<p>
The core client assigns credit for a completed work unit
as a function of the CPU time used,
and the performance metrics of the CPU (discussed later).
This is sent back to the scheduling server;
of course, the number can't be trusted in general.
<p>
Output files may be wrong.
This can happen because of hardware failures,
or because of tampering.
<p>
Both problems - credit-cheating and wrong result -
can be addressed by <b>redundant computing</b>
and <b>result validation</b>.
This means that each workunit is processed at least twice.
The project back end waits until a minimum number
of results have been returned,
then compares the results and decides which are "correct".
The notion of "equality" of results,
and the policy for deciding which are correct,
are project-specific.
<p>
The back end then marks correct results as "validated",
finds the minimum reported credit for the correct
results of a given workunit,
and assigns this amount of credit to all the correct results.
This ensures that as long as a reasonable majority of participants
don't falsify credit, almost all credit accounting will be correct.
<p>
<b>To do</b>: database keeps track of two types of credit:
validated and unvalidated.
Users can see the workunits that didn't pass validation,
or that were given reduced credit.