| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Active Research
Forum: Africa Rainfall Project
Thread: The Extremes thread |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 113
|
|
| Author |
|
|
nekomi_ch
Cruncher Joined: Apr 23, 2024 Post Count: 15 Status: Offline Project Badges: 
|
Sorry for asking but how do you identify an extreme
|
||
|
|
geophi
Advanced Cruncher U.S. Joined: Sep 3, 2007 Post Count: 103 Status: Offline Project Badges: 
|
Sorry for asking but how do you identify an extreme There may be other places but you can find it here . |
||
|
|
Mike.Gibson
Ace Cruncher England Joined: Aug 23, 2007 Post Count: 12373 Status: Offline Project Badges: 
|
Sorry for asking but how do you identify an extreme ARP1 units are different from anything else. Each unit crunches 2 days of weather in 35,609 different patches of Southern Africa. The unit number is the 7 digit part of the name. The 3 digit part is the generation. It corresponds to the 2 day period being crunched. When the unit has been crunched and validated and compared with the actual weather that occurred during that 2 day period. the unit is released back to us as the next generation. Some units have got way ahead of units so some procedures were introduced by WCG to help the laggards catch up. Those within 10 generations of the leaders were classified as "normal", those in the next 5 generations were classified as "accelerated" and the rest as extremes. I also introduced an unofficial classification of "ultra" to refer to 3 units stuck in generations 021 & 022. As far as I can tell, the laggards have been caused by extremes of weather that were outside the parameters of the programming and therefore stuck. Most were restarted by shortening the intervals within the programming called the TimeStep, which is usually 36 seconds. Some have stuck and been restarted several times. Hence they are well behind the leaders. The "normals" are sent out as 2 copies with a 6 day deadline, the "accelerated" also 2 copies but a 3 day deadline and the extremes as 3 copies with a 3 day deadline. Any extra copies sent because of errors or No Reply are sent with the original deadline halved. We have not been informed of the reason for the ultras. Mike |
||
|
|
alanb1951
Veteran Cruncher Joined: Jan 20, 2006 Post Count: 957 Status: Recently Active Project Badges: 
|
Mike,
----------------------------------------Extremes get an initial 1.5 day deadline, or, at least, so do all the ones I ever receive :-) Cheers - Al. [Edit 1 times, last edit by alanb1951 at Feb 11, 2025 9:16:15 PM] |
||
|
|
Mike.Gibson
Ace Cruncher England Joined: Aug 23, 2007 Post Count: 12373 Status: Offline Project Badges:
|
Al
I was relaying what IBM set up. I suspect Krembil modified it for the recent restart. Many of the early ones were resends so would have been halved to 1.5 days. So they may have thought why not do all them all.. Whether they will continue it we will have to wait and see. Mike |
||
|
|
alanb1951
Veteran Cruncher Joined: Jan 20, 2006 Post Count: 957 Status: Recently Active Project Badges:
|
Al Fair enough. No criticism intended -- I just reckoned that people would want to know what's current, and it appears that this has been the case for a long time now so I doubt it'll change. (In my opinion, the shorter deadline actually makes sense anyway!)I was relaying what IBM set up. I suspect Krembil modified it for the recent restart. Many of the early ones were resends so would have been halved to 1.5 days. So they may have thought why not do all them all.. Whether they will continue it we will have to wait and see. Mike I have just checked my wingmen database for ARP1 tasks where the first wingman task receipt and its associated deadline fell into various categories, and I have 1.5-day deadline tasks (all with three initial wingmen) from 2022-02-11 onwards. I can't find a single 3-day task with three initial wingmen during the time I've been collecting data (though I did find one where one wingman failed within 15 minutes, which might've looked like an initial three count on a cursory inspection!) Cheers - Al. |
||
|
|
Mike.Gibson
Ace Cruncher England Joined: Aug 23, 2007 Post Count: 12373 Status: Offline Project Badges:
|
Al
I haven't been storing that data so must concede. However your example would fit my suggestion. I presume that the extremes have all become extreme at some stage as a result of needing resends or even TimeStep changes. Mike |
||
|
|
alanb1951
Veteran Cruncher Joined: Jan 20, 2006 Post Count: 957 Status: Recently Active Project Badges:
|
With the reappearance of lower-generation work-units during 2025-02-11, I got this one...
----------------------------------------ARP1_0034320_125_0 Fedora Valid 2025-02-11T20:49:35 2025-02-12T03:45:26 6.73/6.77 It has moved 4 generations in a week (last sighted by spRocket on the 5th), having moved 4 generations in the previous 5 days. [Edited to acknowledge wingman 2 validating.] I just noticed that this one probably failed Adri's critical "should I post it" test, but we'd been commenting on it recently :-) With that in mind I will point out that ARP1_034248 has just validated at generation 128 (where there are lots of other units), having taken 7 days to advance from generation 123, but offer no details :-) Cheers - Al. [Edit 1 times, last edit by alanb1951 at Feb 12, 2025 1:32:31 PM] |
||
|
|
adriverhoef
Master Cruncher The Netherlands Joined: Apr 3, 2009 Post Count: 2160 Status: Offline Project Badges: 
|
Here's another sighting of a validated workunit in the Extreme category with 5 or less workunits in its current generation (or in both its adjacent generations).
(Output generated by ”wcgstats -frrr= 663882302”) workunit 663882302 ARP1_0034591_124_0 Linux Debian InPrg 2025-02-12T02:49:36 2025-02-13T14:49:36 0.00/0.00 0.0/0.0 Adri |
||
|
|
Mike.Gibson
Ace Cruncher England Joined: Aug 23, 2007 Post Count: 12373 Status: Offline Project Badges:
|
There are currently 347 extremes of which 82 are missing from state.txt and are presumably stuck. This probably includes the 3 ultras and the 8 in generations 104 - 110.
32 moved on a generation in the last 24 hours. Mike |
||
|
|
|