| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Active Research
Forum: Africa Rainfall Project
Thread: Deadline for Africa Rainfall resends too short |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 14
|
|
| Author |
|
|
Crystal Pellet
Veteran Cruncher Joined: May 21, 2008 Post Count: 1403 Status: Offline Project Badges: 
|
The deadline for an ARP1-resend is 58.8 hours. That's the normal 35% of the original 7 days deadline.
However some of my hosts are not or hardly able to return the result before that short deadline of a resend. I suppose not all crunchers have the most state of the art hosts and not all machines are running 24/7 and are using 100% CPU. As an example: An AMD Opteron has a clockspeed of 2.1GHz and still needs > 55 hours elapsed time when running full time 100% for an ARP1-task. Please increase the deadline for ARP1-resends and/or make the deadline of an original task 10 days when the first is not possible. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Second first but not the second (longer original deadline - OD). Every day added to the OD will slow down getting a full step 1. It's not been explained though asked, but I'd not be surprised if adjacent blocks are used as input to generating a subsequent 48 hour set of parameters.
----------------------------------------Could a project scientist [Edit 1 times, last edit by Former Member at Nov 14, 2019 1:41:57 PM] |
||
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7845 Status: Recently Active Project Badges: 
|
Could a project scientist elude on this? I agree, but perhaps a better word would be elucidate, rather than elude. Cheers
Sgt. Joe
*Minnesota Crunchers* |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I agree too. 35% of 10 days is reasonable. 35% of 7 days is pushing it for WUs of this size. (84 vs 58.8 hrs).
However, I expect the techs to want to see the number of aborted re-sends before changing anything, as we've effectively got the additional length of time it takes to process a WU to finish it before it's declared too late. Of course, it also means that there are more WUs in the wild than there might otherwise be ... |
||
|
|
sunfolk
Master Cruncher Super Kiwi Socialistic Empire Of Jacinda Joined: Oct 8, 2006 Post Count: 1769 Status: Offline Project Badges: 
|
Way Too Short.
----------------------------------------It's a bit galling when you return a resend work unit within 2 days and after 25 hours CPU time, when the original work units have been given seven days to complete and take 43 hours, and then have your resend rejected for taking too long, by like an hour....  feels like volunteer abuse     |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Way Too Short. It's a bit galling when you return a resend work unit within 2 days and after 25 hours CPU time, when the original work units have been given seven days to complete and take 43 hours, and then have your resend rejected for taking too long, by like an hour.... feels like volunteer abuse That's a bit puzzling as that resend would have come in well before the next resend. Maybe you can post a copy of the full distribution to see the timings. |
||
|
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges: 
|
Can't the trusted/reliable machines take care of that? I run mine 24/7, keep the default buffer of 0.1/0.5 days, and have a Ryzen 1700 (or faster) on Ubuntu 18.04.
Since they don't send me many, we may never know. |
||
|
|
Crystal Pellet
Veteran Cruncher Joined: May 21, 2008 Post Count: 1403 Status: Offline Project Badges:
|
Can't the trusted/reliable machines take care of that? They do, else they would not get a resend. A reliable machine however can also be a slower machine.When a machine returns (almost) no errors and a lot of successes within 4 days, the machine is reliable and is a candidate for receiving resends even with a shorter deadline than it's normally returning ARP1's. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Has the policy of reliability expired on the part of being known to return on average within 48 hours?
----------------------------------------[Edit 1 times, last edit by Former Member at Nov 17, 2019 2:59:10 PM] |
||
|
|
Crystal Pellet
Veteran Cruncher Joined: May 21, 2008 Post Count: 1403 Status: Offline Project Badges:
|
Has the policy of reliability expired on the part of being known to return on average within 48 hours? Reliability is per application.A machine not returning ARP1-tasks within 48 hours on average, would never be reliable enough for getting ARP1-resends. I have 2 hosts both already had an ARP1-resend, but on average needing > 48 hours elapsed time. Could you point me to WCG-info where that 48 hours is stated. |
||
|
|
|