Page 1 of 1

Idea: Low priority WUs

Posted: Sun Apr 19, 2020 6:04 pm
by TPL
As there have been a lot of new people with computers that can be considered a bit substandard nowadays but willing to donate their idle time, why not to add low priority choise into next client version? I have one laptop too which could be used for long term WUs and actually is used by now but I am considering to stop that because it seems to be only marginal.

There are some WUs that have long timeout. Those WUs could be directed to slower computers.

Sorry if this has been up somewhere before, just slipped my mind.

Of cause it should be named some other way like for older computers or anything but not to underestimate anyone. Science is science no matter how fast it goes. Why not to use all the potencial there is?

Re: Idea: Low priority WUs

Posted: Mon Apr 20, 2020 8:09 pm
by davidcoton
The problem with slow computers is that they hold up a whole run/clone -- the next generation start needs the previous complete return. So the deadline are set to encourage the use of reasonably fast computers. I suspect (but have not checked) that the total compute power donated by computers/GPUs that only marginally meet the deadline is a small percentage of the F@h total, so encouraging more of the slower processors will not contribute a lot even on projects that can take the delay will not be worthwhile.

There is a problem trying to ensure that big CPU WUs are not allocated to small CPUs, but that is not quite the same.

Be careful with "priority" -- all projects have a priority set relative to the science-driven urgency -- this is currently used to prioritize COVID-19.