| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Active Research
Forum: Africa Rainfall Project
Thread: Work Available |
| Member(s) browsing this thread: xensazn |
|
Thread Status: Active Total posts in this thread: 3268
|
|
| Author |
|
|
Mike.Gibson
Ace Cruncher England Joined: Aug 23, 2007 Post Count: 12396 Status: Offline Project Badges: 
|
leloft
ARP have established 35,609 separate plots of land in Sub-Saharan Africa. Each plot is a square 3 KM x 3 KM. Generation 000 computed the weather for the 2-day period starting at 00:00 on 1 July 2018. That was then compared with the actual weather from the Weather Channel. Each checkpoint corresponds to another 6 hours. You would see this if you clicked on the status 'valid' in Results Status. Each plot is independent of the other plots but dependent on the previous generation for that plot. When a unit is validated, the next unit is generated for that same plot - in other words the next generation for that plot. I understand that, for some reason, the 6 units in generation 001 did not compute correctly. They are being investigated to find out what went wrong. The other stragglers had fallen by the wayside for one reason or another. Maybe they came back late and got misplaced. I can only guess. They were picked up in a reconciliation and are now proceeding to catch up by being sent out to so-called 'reliable' machines with half the deadline of current generation units. 'Reliable' machines are those which regularly validate units correctly and within a short period of receiving them. I believe the current criterium is to return within 2 days, but that changes. We are now up to generation 090, so a straggler in generation 045, say, has to be crunched 45 times in order to catch up. Generation 090 covers a 2-day period at the end of December 2018. I hope this covers your questions satisfactorily. Mike |
||
|
|
leloft
Cruncher Joined: Jun 8, 2017 Post Count: 23 Status: Offline Project Badges: 
|
Thank you. Very clear and informative explanation. My x/y/z strategy is now an informed strategy!
Many thanks for your trouble. |
||
|
|
Unixchick
Veteran Cruncher Joined: Apr 16, 2020 Post Count: 973 Status: Offline Project Badges: 
|
I just got this 82 that gives me 4.5 days to finish. not a resend. so I'm guessing 82 is now priority??
ARP1_ 0009794_ 082_ 0-- Darwin 20.6.0 - In Progress 8/19/21 16:03:46 8/24/21 04:03:46 0.00 0.0 / 0.0 |
||
|
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges: 
|
ARP have established 35,609 separate plots of land in Sub-Saharan Africa. Each plot is a square 3 KM x 3 KM. Do you know if they are contiguous plots? I would think a good cross-check would be if the results of one plot are similar to an adjacent plot. If not, something went wrong with the data or calculations. |
||
|
|
pwhidden
Cruncher USA Joined: Nov 17, 2004 Post Count: 32 Status: Offline Project Badges:
|
Jim 1348:
----------------------------------------The project is looking a VERY localized rainfall. From the project page: "However, because rainfall in this area is often localized—sometimes almost at the level of one farm—it's difficult to forecast accurately with satellite data, which show larger weather patterns."  |
||
|
|
Mike.Gibson
Ace Cruncher England Joined: Aug 23, 2007 Post Count: 12396 Status: Offline Project Badges:
|
flensr
I, too, have an i7-3770. It is running 4 ARP, 3 MCM & 1 OPN. All those are readily available but ARP is not necessarily immediately available. I have set those numbers in app_config.xml and the profile system limits to 5/4/2 so there is one spare each. I get a regular supply of each. When you start to ask for ARP it can take up to a day to get your full cache because they are only available intermittently during the day. ARP take about 24 hours on my machine when 4 are running. The recommendation is to not use more than half your threads for ARP because of its system requirements. The other threads can be used for other projects. As the other projects are much shorter units, they enable you to ask for ARP more often to catch up your cache. Mike |
||
|
|
Mike.Gibson
Ace Cruncher England Joined: Aug 23, 2007 Post Count: 12396 Status: Offline Project Badges:
|
Jim1348
There is no need for the cross-check that you suggested because they cross-check with the actual weather that occurred from the Weather Channel. Mike |
||
|
|
Mike.Gibson
Ace Cruncher England Joined: Aug 23, 2007 Post Count: 12396 Status: Offline Project Badges:
|
Unixchick
Thank you. As it has been an extra day since you posted, I will wait until Kevin's next post to update progress. Mike |
||
|
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges:
|
I am running 50% of the cores (that is, six cores) on my Ryzen 3600 on ARP, and nothing else. They are doing 9 hours, 50 minutes (Ubuntu 20.04.2).
That leaves me plenty of CPU power free to support a GPU on Einstein, where it helps, without using much cache at all. |
||
|
|
Mike.Gibson
Ace Cruncher England Joined: Aug 23, 2007 Post Count: 12396 Status: Offline Project Badges:
|
Jim1348
----------------------------------------I am running Einstein on my Intel GPU - couldn't get anything but errors on OPNG. I find that by restricting Einstein to GPU on BRP4, I can run all 8 of my CPU threads for WCG. 4 ARP, 3 MCM & 1 OPN. Other GPU jobs on Einstein would use a whole CPU thread. Mike [Edit 1 times, last edit by Mike.Gibson at Aug 20, 2021 5:34:00 PM] |
||
|
|
|