A few notes. 1) The resource share for what to run next needs to be responsive to what is happening on the local computer, and so would have to be based on resource usage. 2) The long term resource share would have to be calculated well after the task is completed - after the credit is granted. This will be anywhere from seconds to weeks after the work is done. 3) How do we handle failed / late / invalid tasks? Ones that have used resources, but are not going to generate credit? Projects with a large % of failed / late / invalid tasks on this compute?
-----Original Message----- From: boinc_dev [mailto:[email protected]] On Behalf Of Eric Korpela Sent: Monday, August 15, 2016 2:37 PM To: CM <[email protected]> Cc: BOINC Developers Mailing List <[email protected]> Subject: Re: [boinc_dev] 4th Gen BOINC credit system There are a few major issues with the current credit system (credit_new). The first is that nobody uses it. Most projects (possibly all but SETI@home, if what I'm told is correct) stayed with the old system. The second is that there is still no consistency between projects, and that wouldn't be fixed even if everyone used credit_new. Which gets us to the third issue. There is no longer any significant cross project relationship between credit and work done. The relationship was tenuous a decade ago when SETI@home was the only project that actually calculated how many FLOPs were being done, now it's nearly entirely gone. The people who volunteer only to boost their RAC will happily migrate to the project with the highest credit grants and the FLOPs shown at stats sites are likely to have no relationship to reality. There are ways that the system could be stabilized. One way would be to tie resource shares to RAC rather than wall time. (I was involved in a project with a 4% resource share. It would typically supply 50% of my total RAC, because its credit grants were in now way tied to actual work done.) If resource share was time to RAC, a project with 4% resourcre share would create 4% of the RAC. If a project wanted to increase the number of participant, it could increase the credit it offered, but that would reduce the net resource share from people who participate in multiple projects. Currently there is no penalty for increasing the credit offered. On Sun, Aug 14, 2016 at 9:06 AM, CM <[email protected]> wrote: > Hey, > > A '4th gen BOINC credit system' thread was created over at the official > BOINC forums: https://boinc.berkeley.edu/dev/forum_thread.php?id=10953 > > There's been about 77 posts to the thread, the general consensus was that > a new BOINC credit system is a good idea but thus far the definition/specs > of such a next gen system has not been agreed upon. > > What are the BOINC dev's thoughts on this topic? I'd love to continue this > discussion & work towards a next-gen BOINC system. > > Cheers guys :) > > _______________________________________________ > 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. > -- Eric Korpela [email protected] AST:7731^29u18e3 _______________________________________________ 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. _______________________________________________ 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.
