+1

Another Sunday, so another full report. This one will be the last before shutdown

There have been no validations reported for the last 2 days although there have been a few results returned - 3 yesterday! maybe they were for units already validated.

The 3 'new' ultras need to advance by 5 generations per week to catch up to the calculated completion date. They are currently in generations 008, 010 & 011 so have not progressed this week.

Only the last of the 'old' ultras is still discernible, now in generation 093. The rest are mixed in with the other unstuck units.

There are 117 units in the extreme range, all of which are listed by WCG.

The definitions of extreme, accelerated(priority) & normal have remained at 125, 135 & 120, respectively.

3,472 units have validated in the week (down to 496.0 per day) so we now have 1,953,789 remaining until the end of the project. We cracked the 2 million mark.

My calculated completion date has been suspended due to the wind down. I now expect the project to end early next year, depending on how many crunchers return after the 2 month break.

Enjoy your holidays!

Mike

I suspect the best way to free up some of the work-load on "faster machines" might be to reduce the normal deadline to five or six days (with or without a grace day) rather than the existing 7+1. Yes, it would upset certain users (not all of them with small/slow systems!) but there are probably quite a lot of retries out there most of the time, and not all because of less powerful systems...

Under normal running conditions my machines were getting 25% or more tasks that had wingmen responding 6.5 days or more after receiving tasks (and the response was often No Reply rather than Error - Not started by deadline). And the vast majority of those late/non-existent replies came from a very small set of systems that never seemed to return work for any project in a timely fashion(*)! Once we got onto clean-up, well over 50% of my retries were for that small subset of devices!

Of course, if some way could be found of making sure that the maximum "upper limit" on ARP1 tasks could be set to [say] twice the number of announced CPUs a lot of the late (and non-existent) returns would go away!

Cheers - Al.

(*) That's based on an analysis of wingman responses for tasks I've run on projects during late January 2022 and February 2022 up to the shutdown...

I use windows and didn't see that sort of problem until the wind down. The problem I saw then could have stemmed from crunchers loading up their machines before the wind down and then having the deadlines cut from 8 to 6 days.

Restricting machines to total threads would be a big help, but the overloaders would not get the extremes or accelerated units as they would not be considered reliable.

The plus 1 day was not a good idea as it encouraged people to return slowly. However, it didn't matter too much as they would only be crunching normal units.

Mike

I'm so excited to have ARPs again. I have 2 and they are both 125. Does anyone know if the script telling us where all the WUs are is still running? I'm feeling lazy and don't want to hunt back through this thread for the link to the scripts. If it is running and you have the links handy can someone post them again??