good thread buise ...this should encourage folders to fold 24/7.
0
Straight_ManGeeky, in my own wayNaples, FLIcrontian
edited February 2004
That stat will also let Folding not only do the fastest for folding overall, it will let those boxes that fold tinkers poorly NOT get Tinkers if they can fold Gromacs better and things are tracked by kind of WU also. In fact, it is partly that the dang servers, especially the ones with new Gromacs to accept and give out, have been overloaded more than the Tinker and older Gromacs WU servers.
Folding needs that kind of info to determine how many new Gromacs to gen, and if andd when to phase Tinkers gradually out, as well as get the large volumes of results that are needed in fastest manner.
I saw some things in that thread that talked about 1 GB boxes, those boxes will get fewer Tinkers as the Tinkers are phased gradually out, and they will be if there are nmore boxes that can fold Gromacs better than Tinkers by a large proportion. Not only is that likely to happen as more faster boxes come online, but we are likely to see the 1 GB boxes gradually get smaller Gromacs as things develop over the long run-- instead of Tinkers. One side effect of having lots of faster boxes folding, is that Folding central ops at Stanford has to ramp up server capability to handle bigger uploads and downloads at a greater bandwidht per day as they get more complex results in faster.
Since Gromacs are being used more and more, it is likely that Folding has some limited funds for this. But they have to conserve bandwidth per good (read definitive, a set of failing WUs on many boxes says that the experiment implied by the one part of the project that fails can be dropped unless other projects later find it-- more likely one variable in it-- needs to be looked at again, and knowing what NOT to persue helps to isoltate down to what works also) result also, so giving boxes what they can best do by setting assignment thresholds based on actual work and how fast they accurately turn in useable results makes sense-- and a floating standard based on time actual versus turnin deadline makes sense. Easiest way to do that is a ratio in percentage or decimal to right of decimal point, and in essence that is what this thing is.
I posted a suggestion that would let the faster computers get the WUs that want doen right away and still allow for people who dont have their machines on all the time get gromacs.
And dont get all fluestered either. My Celeron with SSE at 788MHz constantly gets gromacs. Doesnt take much to get it done with 80% left.
BuiesCreek847
In a van, down by the river, NC Icrontian
Comments
Though been away from this community so long, and lost all interest as to what the heck I'm folding, I just am
Folding needs that kind of info to determine how many new Gromacs to gen, and if andd when to phase Tinkers gradually out, as well as get the large volumes of results that are needed in fastest manner.
I saw some things in that thread that talked about 1 GB boxes, those boxes will get fewer Tinkers as the Tinkers are phased gradually out, and they will be if there are nmore boxes that can fold Gromacs better than Tinkers by a large proportion. Not only is that likely to happen as more faster boxes come online, but we are likely to see the 1 GB boxes gradually get smaller Gromacs as things develop over the long run-- instead of Tinkers. One side effect of having lots of faster boxes folding, is that Folding central ops at Stanford has to ramp up server capability to handle bigger uploads and downloads at a greater bandwidht per day as they get more complex results in faster.
Since Gromacs are being used more and more, it is likely that Folding has some limited funds for this. But they have to conserve bandwidth per good (read definitive, a set of failing WUs on many boxes says that the experiment implied by the one part of the project that fails can be dropped unless other projects later find it-- more likely one variable in it-- needs to be looked at again, and knowing what NOT to persue helps to isoltate down to what works also) result also, so giving boxes what they can best do by setting assignment thresholds based on actual work and how fast they accurately turn in useable results makes sense-- and a floating standard based on time actual versus turnin deadline makes sense. Easiest way to do that is a ratio in percentage or decimal to right of decimal point, and in essence that is what this thing is.
John D.
And dont get all fluestered either. My Celeron with SSE at 788MHz constantly gets gromacs. Doesnt take much to get it done with 80% left.