| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Active Research
Forum: Africa Rainfall Project
Thread: 3.5 day deadline sent to unsuitable machine |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 25
|
|
| Author |
|
|
Mike.Gibson
Ace Cruncher England Joined: Aug 23, 2007 Post Count: 12372 Status: Offline Project Badges: 
|
My old laptop received a re-send because someone was late returning their unit. This regularly takes a long time to crunch arp and actually took 103 hours elapsed (87 hours CPU). Meanwhile the original late unit was returned so my unit was declared 'Too Late'.
This is a waste of my crunching time when I could have been crunching something else. My machine should never have been sent a 3.5 day unit. Mike |
||
|
|
yoerik
Senior Cruncher Canada Joined: Mar 24, 2020 Post Count: 413 Status: Offline Project Badges: 
|
tbh, that's why I don't run ARP on my laptops anymore. They take too long to do, to justify the resource cost on my end.
----------------------------------------There isn't much you can do, other than abort one the second you see it or stop crunching ARP on that device.  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Untick ARP1 in the device profile and also "if there is no work for.... send me something else". Then they will never arrive, not even through the back door.
|
||
|
|
BladeD
Ace Cruncher USA Joined: Nov 17, 2004 Post Count: 28976 Status: Offline Project Badges: 
|
Yep, my oldest/slowest PCs are also unattended and I don't run ARP WUs on them.
----------------------------------------But I will have check my results status when the update finish.  Edit: No WUs here with a Too late status. ---------------------------------------- [Edit 1 times, last edit by BladeD at Sep 28, 2020 12:50:10 AM] |
||
|
|
Mike.Gibson
Ace Cruncher England Joined: Aug 23, 2007 Post Count: 12372 Status: Offline Project Badges:
|
I was quite happy crunching 7 day units and thought that the 3.5 day units were only supposed to go to those machines deemed 'reliable' otherwise known as quicker.
Mike |
||
|
|
yoerik
Senior Cruncher Canada Joined: Mar 24, 2020 Post Count: 413 Status: Offline Project Badges:
|
reliable just means that you are returning good results, from my understanding. Nothing to do with the speed -- so long as they return on time.
---------------------------------------- |
||
|
|
Mike.Gibson
Ace Cruncher England Joined: Aug 23, 2007 Post Count: 12372 Status: Offline Project Badges:
|
Reliable also means regularly returning units within half the deadline.
Mike |
||
|
|
Link64
Advanced Cruncher Joined: Feb 19, 2021 Post Count: 129 Status: Offline Project Badges: 
|
Reliable also means regularly returning units within half the deadline. Yes, but apparently this is determined across all WCG projects. I've tried ARP now and I have the impression, that 3.5 days deadline causes lots of wasted CPU time, here an example: ARP1_ 0019494_ 060_ 4-- In Progress 4/30/21 04:07:08 5/3/21 16:07:08 0.00 0.0 / 0.0 Seems like lots of computers are not able to return the results in such short time, the _1 was also returned after deadline and my computer will return _3 in 3 hours or so. It needs 35+ CPU hours for one ARP task and runs about 12 hours a day, so no idea why it was assigned such WU at all, those should only go to computers running nearly 24/7 or with a lot faster CPUs than my C2D. WUs with 7 days deadline run however fine on it so far.  |
||
|
|
Mike.Gibson
Ace Cruncher England Joined: Aug 23, 2007 Post Count: 12372 Status: Offline Project Badges:
|
Link64
I started this thread because a short unit was sent to my laptop which takes 4 days to crunch without considering queueing time. What you say is right. However, the main problem is those people with slower machines who have too long a queue. If most of time is wasted on queueing, there isn't much left for crunching. Initially, there wer not enough units to go round so many crunchers allowed longer queues to tide them over. Now that the units are in plentiful supply, there is no reason to hold a large cache. My PC crunches an ARP unit in about 23 hours and I have it set to crunch a maximum of 2 with a total of 3 held, so only 1 spare. That is enough. Mike |
||
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7666 Status: Offline Project Badges: 
|
Years ago there were requests for the information about cpu's to be incorporated into the scheduling algorithms for the servers. With enough information about the cpu speed and the average time from when the machine requested a work unit to when it returned a completed work unit, resends could be sent to the machines most likely to return the unit before the new deadline. To the best of my knowledge this scheme has never been implemented here at WCG. I believe reliable is a designation about returning so many units on time and valid. I believe the number of units meeting this criteria is fairly small, perhaps under 10. Thus a machine can be determined to be reliable in a fairly short time.
----------------------------------------If any others have additional info, please chime in. Cheers
Sgt. Joe
*Minnesota Crunchers* |
||
|
|
|