| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: Help Cure Muscular Dystrophy - Phase 2 Forum
Thread: Pseudo very short WUs |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 8
|
|
| Author |
|
|
JmBoullier
Former Community Advisor Normandy - France Joined: Jan 26, 2007 Post Count: 3715 Status: Offline Project Badges: 
|
After the series of really very short WUs which were sized as the normal short ones I am now receiving a new series of WUs sized like the former very short ones, but actually the one I am testing is gone for twice longer than the normal short ones.
----------------------------------------Confused? OK, see examples below with all times based on the same machine/OS: 1. Normal short WUs, the vast majority so far: - avg duration 55-60 minutes - initially estimated as other projects at the beginning, stabilized at about 1 hour now 2. Very short WUs of the GPDAA/RADIA series - avg duration 8 minutes - sized as the normal short ones 3. New pseudo very short WUs in the 2O72/2O72 series - projected time of one WU under test: 2.30 hours - initially estimated at 9 minutes. I had received another single WU (alone of its kind) in the GPDAA/MYH14 series like those bad ones. It has just finished in 50 minutes and has already messed up all duration estimates in this machine. When the new ones will complete at more than two hours it will be a total mess for days... Please correct those sizings and abort the pending ones from the servers. Jean. |
||
|
|
nasher
Veteran Cruncher USA Joined: Dec 2, 2005 Post Count: 1422 Status: Offline Project Badges: 
|
honestly i like the short ones ... well semi short...
----------------------------------------Right now im trying to clear some other projects from my other computers to get them ramped up for the dual badge run but i only have 4 running machines and not many dual cores and no more than dual at all so its hard to compete with some of you on the quad cores and such but im doing all i can and makeing my part personaly any Wu under 4 hours is nice in my opinion any WU over 24 hours gets anoying to me  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
honestly i like the short ones ... well semi short... Right now im trying to clear some other projects from my other computers to get them ramped up for the dual badge run but i only have 4 running machines and not many dual cores and no more than dual at all so its hard to compete with some of you on the quad cores and such but im doing all i can and makeing my part personaly any Wu under 4 hours is nice in my opinion any WU over 24 hours gets anoying to me Don't feel overwhelmed. I'm only running a P4 HT and still trying to catch up with Clean Energy. It's my server so it is on 24/7 but has been chugging along pretty well for the past year. I still have CMD2 and Influenza which I haven't even started. [Edit 1 times, last edit by Former Member at May 13, 2009 2:51:03 PM] |
||
|
|
JmBoullier
Former Community Advisor Normandy - France Joined: Jan 26, 2007 Post Count: 3715 Status: Offline Project Badges:
|
The purpose of this thread is not about very short WUs, I like them too from time to time and I enjoyed looking at the GPDAA/RADIA series of very short WUs being crunched a few hours ago.
----------------------------------------Now we have WUs which pretend to be as short as those ones while, in fact, they will need 15 times more runtime than what they announce. The first visible and very annoying effect is that as soon as one of these false very short WUs is finished the estimated duration of all jobs in the queue is raised considerably and the whole queue is running in high priority mode despite the comfortable 14-day deadline. Cheers. Jean. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I hope this time estimate works out OK for all crunchers. I have 900+ HCMD2 WU In Progress, but they all appear to be similar to
CMD2_ 0001-GPDAA.clustersOccur-MYH3.clustersOccur_ 1248_ 0-- |
||
|
|
roundup
Veteran Cruncher Switzerland Joined: Jul 25, 2006 Post Count: 831 Status: Offline Project Badges:
|
The first visible and very annoying effect is that as soon as one of these false very short WUs is finished the estimated duration of all jobs in the queue is raised considerably and the whole queue is running in high priority mode despite the comfortable 14-day deadline. Cheers. Jean. Same effect here. I set the work buffer to 1,69 days because I had to disconnect the Quad from the internet for some hours - expecting that the 1,69 days are crunched in half a day due to the experiences with the GPDAA/RADIA units. The Computer downloaded about 210 WU's with an indicated duration of 10 min 31 sec. After completion of the first WU all remaining WU's changed the estimated duration from 10,5 minutes to 1,75 hours :-(. I got 190 WU's left and will try to crunch them all B-) [Edit 1 times, last edit by roundup at May 13, 2009 4:17:55 PM] |
||
|
|
AnRM
Advanced Cruncher Canada Joined: Nov 17, 2004 Post Count: 102 Status: Offline Project Badges:
|
Perhaps more time in BETA would have been instructive? There may be a good reason for doing it this way but it's obviously causing problems.
---------------------------------------- [Edit 1 times, last edit by AnRM at May 13, 2009 5:35:45 PM] |
||
|
|
JmBoullier
Former Community Advisor Normandy - France Joined: Jan 26, 2007 Post Count: 3715 Status: Offline Project Badges:
|
Complement to my opening post.
----------------------------------------It seems that I have not been lucky when I have chosen a WU to test among these new very short WUs. I have decided to activate all of them before the rest of the queue and it appears that the majority is really very short WUs, although a little longer than previously (20 minutes in average, vs 9 minutes last night and this morning). So far I have found one other big one lost among this pool. I have also got a bunch of similar WUs in my P4 but these ones have names belonging more or less to any other series. So the problem of the client's scheduler being blown up when finding a false very short WU remains, but the number of occurrences is far below what I was fearing. Cheers. Jean. |
||
|
|
|