| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Active Research
Forum: Africa Rainfall Project
Thread: Not enough time given for a task |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 102
|
|
| Author |
|
|
blyons123
Cruncher Joined: Jan 2, 2007 Post Count: 9 Status: Offline Project Badges: 
|
It doesn't make sense to give a deadline of only one week for an 18-hour task the same as a two hour task in other projects !?
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
That same comment came from someone else a week or so ago, but no 18 hours crunching on 7 day deadline is still only 2.5 hours a day, were it not that ARP1 does checkpoint only once per 12.5% progress so you have to run BOINC at least 3 hours uninterrupted to reach the next save point.
Why 7 days? The next step depends on the previous result i.e. unitl you finish your result and report there wont be a next task to send out. There's 180 48 hour simulations in a sequence. If everyone would return the result by the maximum allowed time it would take 1260 days (7x180) to do a full 1 year simulation. Too long. Can't do that, than don't opt in. |
||
|
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1672 Status: Offline Project Badges: 
|
Additionally, for me the question is more "Is it meaningful to give a 7 or 10 days deadline for a 4 or 6 hours task?".
----------------------------------------Initially the 10 days deadline came from a time of slow single core processor; considering that the members would only compute a couple of hours per day. Today, we have in many cases powerful multicore CPUs and many contributors participate on an 24/7 basis. The initial deadline on 7 or 10 days has been kept with respect to the members with less powerful machines, not for encouraging members to maintain a 7 day large buffer, always reporting at deadline and sometime too late. Lavaflow is fully correct with the results dependency. For projects with very large data sets (such as ARP1), the additional problem is the required storage space: as longer the deadline, as larger the needed space for keeping the WUs available until the complete result validation is achieved. A 24 hours deadline as for FAH2 caused me some troubles in the past, because, at this time, I used to maintain a 2.5 days buffer. Today with a much better internet connection (FO), I reduce the whole buffer to 1.25 days and I fine tune the machines with app_config.xml. On my side, the standard deadline could be limited to 5 days. Cheers, Yves |
||
|
|
rbotterb
Senior Cruncher United States Joined: Jul 21, 2005 Post Count: 401 Status: Offline Project Badges: 
|
I can deal with the 18 hour ARP WUs, but I wished the checkpoints were more often than every 12.5% (say more like every 5%). I'm now getting close working on a new job that will not allow me to run my laptop as long as I'm doing now each day. With the checkpoints every 12.5% increment, while I could still get about 50% done on the weekends, I would not be able to run Monday-Friday long enough to get to another checkpoint on weeknight evenings and I won't run my laptop later at night due to all the potential hackers out there.
I have one more ARP WU now running to finish and get my Gold badge. After that, I guess I'll have to stop with this project until I see more checkpoints being added these WUs. I've run into this issue with other projects in the past, and generally once more checkpoints get added, I'm then able to return down the road. While I would love to continue to contribute to this project, I guess for now I'll have to settle for my Gold badge worth of work. Every little bit helps more science forward to new solutions. |
||
|
|
ca05065
Senior Cruncher Joined: Dec 4, 2007 Post Count: 325 Status: Offline Project Badges: 
|
If you hibernate or sleep your laptop instead of shutting down then when you switch it on again the tasks will continue from when the hibernate / sleep was requested.
|
||
|
|
yoerik
Senior Cruncher Canada Joined: Mar 24, 2020 Post Count: 413 Status: Offline Project Badges:
|
the project's needs dictate the deadline.
----------------------------------------Ex: Fight Aids at Home has 24 hour deadlines, because their results build on each other. Not the best way to put it, but I think you know what I mean.  |
||
|
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1672 Status: Offline Project Badges:
|
Hi rbotterb,
----------------------------------------at this time, there is no chance for additional checkpoints. It was comprehensively explained at project start. Cheers, Yves |
||
|
|
rbotterb
Senior Cruncher United States Joined: Jul 21, 2005 Post Count: 401 Status: Offline Project Badges:
|
Fair enough. I'll be finishing my last ARP WU today to get my gold badge and then move on with other projects.
|
||
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7662 Status: Offline Project Badges:
|
I fired up my Q6600 for a couple of days just intending to get OPN units. OOPS, I forgot to change the profile for that machine. It downloaded an ARp unit too. I decided to just let it run. It took a while.
----------------------------------------ARP1_ 0022936_ 019_ 1-- Valid 8/11/20 02:53:29 8/16/20 05:14:51 80.45 / 80.70 1,387.3 / 950.5 Just a little over 80 hours. Not recommended for this machine, but it did complete in time. It took about 5 days, roughly about 16 hours of running per day. That is one of my few non 24/7 machines. After that marathon it will be given a rest for awhile as it is just not very efficient. Just thought I would put that information out there for anyone else wih a fairly low powered machine. Cheers
Sgt. Joe
*Minnesota Crunchers* |
||
|
|
Mike.Gibson
Ace Cruncher England Joined: Aug 23, 2007 Post Count: 12360 Status: Offline Project Badges:
|
ARP units correspond to a particular patch of sub-saharan Africa. The fist number is the number of the patch (about 35,500 of them) each 3 KM x 3 KM. The second number is the 2-day period starting from 1 July 2018 and we are crunching a whole year for the project.
If the deadlines were not kept tight, the patches would get out of step with each other, Generally with WCG, if re-sends go out after the halfway point of a deadline, the time allowed is only half the standard time, so they only go to those machines which are considered 'reliable', which effectively is those which regularly return units within 3.5 days. There is also the aspect that there is a long period between checkpoints and whenever a machine is shut down it loses any time after the last checkpoint. That can be overcome by putting the machine into 'sleep' or 'hibernate' mode, but really the project is only suitable for machines that operate 24/7. As we have only done about 10% so far, any lengthening of the deadline would mean the project taking even longer than it currently will. Mike |
||
|
|
|