Back in the days of the 100,000 step jobs, the time allowance was, ,if I recall correctly, 4 days. Even then I'd get the odd retry after someone had failed to return a job on time...

Now, we have two batches of data being run through which are, as tar as I am aware, identical except for the number of steps per task - one lot does 10,000 steps and the other 50,000 steps.

A 2-day limit might seem reasonable for the 50,000-step jobs, but would be inappropriate for the the 10,000-step jobs as it would take up to 5 times as long to achieve the same total number of steps

I think I saw something about there being 2.5 million steps in total (though I've been unable to find said reference, if it ever existed outside my imagination) -- and the highest iteration number I've seen in a 50,000 step job so far is 044 so that's not disproved yet. If it is right, the shorter jobs will need 250 task iterations to get the required number of steps, and I don't think it requires much imagination to work out what would happen if people with mixed workloads and largish buffers could spend 2 days on an 10,000 step task!

So Sgt. Joe (as quoted by Shaky Jake recently) has the right of it, I would think. They've already added a limit on the number of concurrent FAH2 tasks one can have as part of an effort to stop overruns or contention with other projects, and apart from putting different time limits on the two different types of job I don't see what else they can actually do if they want the results in a timely fashion!

And whilst I do appreciate that those limits aren't encouraging to crunchers with slow machines, it doesn't make sense to hamstring a project just so under-resourced systems can have a go (witness the warnings about CEP2...) My ancient Core 2 Duo Macbook Pro (clocked at about 2GHz) would take 14 to 16 hours to chew through a 50,000 step job (having taken about 28 hurs on the two 100,000 step jobs it got sent without my say-so) so I excluded FAH2 from the list of jobs it could have (just as I had previously not let it get CEP2 tasks) Sometimes, a little common sense goes a long way!

Cheers - Al.

I run FAH2 along with all other projects on a Ryzen 1700 with the default buffer of 0.1+0.5 days and don't have a problem.
I sometimes see the FAH2 running high priority, but that is about it.

One of the machines on which I run WCG is both fast and modern, and like all my machines it runs with a minimal buffer (my other WCG PC host is older, but still very much usable). However, FAH2, with its absurdly short deadline, gave me endless problems with disruption of other BOINC projects, and missed deadlines. And despite what is being said here, I often found myself with multiple FAH2 tasks scrambling over each other and other projects, desperately trying to meet the deadline. Sometimes I would only be a couple of hours over, but by then the task had already been sent to another host. I gave up, and dumped FAH2. Good luck with all those unmanaged hosts that accept tasks and rarely if ever do anything useful.

BOINC was set up to use spare CPU cycles donated by volunteers on millions of ordinary computing platforms around the world. The FAH2 project should either work within the BOINC paradigm, just like lots of other projects do, or buy themselves their own full-time supercomputer. As it is, IMHO they are abusing BOINC.

I believe this subject has been pretty thoroughly vetted. The scientists have said this is the model which works best for them to get heir results in a timely fashion. I have run this project in the past (not currently running it now) and it is my opinion that this project runs best when it is the only project running on a machine. And, it runs by design with a short queue ( loosened a slight bit lately per Lavaflow's request). If this project is disrupting your other projects, you could limit it to only using a specified number of cores, leaving some cores available for your GPU's. The bottom line is, that if this project does not work for you in your environment, don't run it. After all, it is your resources you are donating to various causes and these projects should not cause you undue stress.
Cheers

Was already explained a few days ago and repeated ad nauseam. The tasks have a dependency and a sequence of hundreds. One result forms the base for generating the next step set. With 24 hours deadline and a sequence of 300, they know they'll have a complete simulation series in at most 300 days for that target. If they'd allow the common 7-10 days deadline, they'd not have a simulation complete until 2025.

Enough said.