On Sep 25, 2009, at 7:41 AM, [email protected] wrote:
> In other words, you propose penalizing people that use their
> machines the
> way that BOINC is designed to be run - multi project and offline for
> much
> of the time.
>
> David Anderson is trying to get people to do multiple projects, and
> some of
> the proposals run directly counter to that. Penalizing people for
> using
> the program the way it is designed and promoted to work is wrong.
If he is he has a funny way of doing that ... most of the changes to
the Resourcce Scheduler in these last few months has been a direct
slap in the asperations of those that do want to run many projects.
And I qouote:
• Work fetch is limited to ensure that deadlines can be met. A
computer attached to 10 projects might have work for only a few
(perhaps only one) at a given time.
• If deadlines are threatened, the CPU scheduling policy optimizes
the likelihood of meeting deadlines, at the expense of variety.
http://boinc.berkeley.edu/trac/wiki/ClientSched
NOTE: ... (perhaps only one) ...
NOTE: ... at the expense of variety.
Yes sir! Run out there and sign up for a variety of projects and we
will do our best to ignore your wishes ...
I will also point out that since 6.6.20 (or there about) the Work
Fetch has been seriously bent ... I know what my resource shares are
and what my run times are and about how much work should be pulled and
in no case should BOINC pull 6 plus CPDN tasks at one time (I still
have 4 to go) ... but my suggestion on how to fix that was
rejected ... a fix supported by CPDN I might add ...
I wrote about this trend before mid-year and documented it as well as
I could and was ignored.
I wrote about it again this last week in the specific context of GPU
scheduling and resourcing ... and to this point ... have been ignored.
But it is not going unnoticed by others who are just as unhappy about
it as I ... though they are likely smarter because they are not
bothering to try to get it changed ... recognizing that it is likely a
waste of time to do so ...
But you cannot seriously assert that we are struggling to get people
to run a broader variety of projects when the direction of change is
to make that less rewarding ... I will point out that the long
lingering cross-project credit parity issue does not help either ...
and the weak proposal suggested is not going to do much to change it
either ... and when I asked if this was a serious proposal and it user
input would be listened to I have gotten no answer ...
On the other hand, credit bonuses as awarded by GPU Grid did pull in
some people ... and get some to upgrade their GPUs to get the bonus.
_______________________________________________
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.
