| 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: t1947m |
|
Thread Status: Active Total posts in this thread: 3308
|
|
| Author |
|
|
Mike.Gibson
Ace Cruncher England Joined: Aug 23, 2007 Post Count: 12432 Status: Recently Active Project Badges: 
|
My re-send, which I returned yesterday, was followed by the original wingman 11 hours after me. He took 8.5 days to return something that he only crunched for 28 hours. Either he only operates part-time, in which case he shouldn't be running arp or he had much too large a cache, which should be prevented.
All 3 were given credits. Mike |
||
|
|
Jean-David Beyer
Senior Cruncher USA Joined: Oct 2, 2007 Post Count: 338 Status: Offline Project Badges: 
|
[OT] If you're interested I am using a Ryzen 9 3900 X stock speed. And I run about 14 hours a day [/OT] I now have a DELL 5820 with a 16-core (count 'em: 8 hyperthreaded ones) Xeon processor. 32 GBytes RAM, 512 GByte SSD, and two SATA hard drives. Currently running Red Hat Enterprise Linux 8.2). I normally let it run 24/7.  |
||
|
|
Speedy51
Veteran Cruncher New Zealand Joined: Nov 4, 2005 Post Count: 1297 Status: Offline Project Badges:
|
[OT] If you're interested I am using a Ryzen 9 3900 X stock speed. And I run about 14 hours a day [/OT] I now have a DELL 5820 with a 16-core (count 'em: 8 hyperthreaded ones) Xeon processor. 32 GBytes RAM, 512 GByte SSD, and two SATA hard drives. Currently running Red Hat Enterprise Linux 8.2). I normally let it run 24/7. Thanks for details. Funny thing we are running the same amount of RAM I am running Windows 10 professional. We also have the same size SSD.  |
||
|
|
alanb1951
Veteran Cruncher Joined: Jan 20, 2006 Post Count: 977 Status: Offline Project Badges:
|
@aurum420
----------------------------------------I started clicking on the 2s I completed and they all look like this: Result Name OS type OS version App Version Number Status Sent Time Time Due / People are getting these serial WUs and sitting on them until they expire after 7 days. They should be banned. Seven days is too long, should be 3 days.Return Time CPU Time / Elapsed Time (hours) Claimed/ Granted BOINC Credit ARP1_ 0020641_ 034_ 2-- Linux Linuxmint Linux Mint 20 [5.4.0-53-generic|libc 2.31 (Ubuntu GLIBC 2.31-0ubuntu9.1)] 727 Valid 11/17/20 08:42:58 11/18/20 07:41:36 11.38 231.4 / 415.9 ARP1_ 0020641_ 034_ 0-- Linux 3.10.0-1127.18.2.el7.x86_64 727 Valid 11/10/20 08:38:10 11/11/20 07:10:00 16.34 600.4 / 415.9 ARP1_ 0020641_ 034_ 1-- Linux 4.4.0-87-generic - No Reply 11/10/20 08:37:49 11/17/20 08:37:49 0.00 0.0 / 0.0 The scoring makes no sense either. _0 16.34 600.4 _1 0 0 _2 11.38 231.4 I ran it 30% faster, ran it the same day I got it, and 64% fewer points. That's gratitude for ya. I get the occasional _2 or _3, but a lot of the time they end up Server Aborted (so I've just wasted the bandwidth - fortunately, I've got decent broadband!,,,) If everyone who did WCG stuff only did WCG stuff I suspect we'd see a lot fewer jobs miss deadlines, but that's just speculation! Also, bear in mind that many WCG active users are probably "install it and leave it to it" users, or may be using Science United (in which case they have little or no control over how things run beyond shrinking queues in the BOINC client!) Whilst one might be annoyed about it, one has to accept it, I think...  As for the "scoring" - it is actually consistent with the way they handle differing claims; sum the two claims and divide by two... Sadly, it looks as if your wingman on that job needs to do a re-benchmark to get the claims sorted out!!! For what it's worth, your claim of around 600 is probably about right (my claims seem to be between 560 and 650 on both an i7-7700K (taking 11 or 12 hours) and a Ryzen 3700X (taking 9.5 to 11 hours). Fortunately, as you'll always have a wingman for ARP1, it'll balance out over time! For tasks that may not always have a wingman it can be problematic if your claim level drops, because you'll just keep getting what you ask for (provided the server thinks it looks reasonable!) until you get a wingman again, and if you're asking low... Just keep crunching... Cheers - Al. [Edited to fix a typo...] [Edit 2 times, last edit by alanb1951 at Nov 19, 2020 4:31:03 AM] |
||
|
|
Mike.Gibson
Ace Cruncher England Joined: Aug 23, 2007 Post Count: 12432 Status: Recently Active Project Badges:
|
MIP & SCC are the only projects with a 10 day deadline and their units are readily available so the first step should be a reduction of the maximum cache to 7 days.
If it were possible, this should be adjusted according to average reporting over the last week or month. A download limit of 22 for a machine with 32 threads would not be sufficient. It would need to be at least 1.5 per thread to keep running during up- and down-loading. The current 64 limit would be more than enough, so scrap the "unlimited". Any other changes would impact on the larger farms operating 24/7. Mike |
||
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7694 Status: Recently Active Project Badges: 
|
This could easily be fixed by first eliminating the Unlimited setting in Device Profiles The problem is not with the "unlimited" but with the cache setting. If some are using cache settings in excess of 5 days with ARP and they are not running 24/7 there could be a problem. The problem would disappear with a cache setting of 1 day which should be sufficient for most people as the supply of most projects is quite consistent on WCG. Outages are few and far between. Cheers
Sgt. Joe
*Minnesota Crunchers* |
||
|
|
Mike.Gibson
Ace Cruncher England Joined: Aug 23, 2007 Post Count: 12432 Status: Recently Active Project Badges:
|
Cache settings really are the problem, but reducing it to 1 day is excessive bearing in mind that a lot of people take more than 1 day even operating 24/7. I know that I said 7 days as a first step, but that was so as not to offend too many people. A more realistic limit would be 3 days, but that could be done in 2 or 3 stages.
3 days would allow for anyone operating 50% of time crunching. It would also allow for most WCG outages which have occurred several times this year on ARP, although not on other projects. Mike |
||
|
|
|