| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: The Clean Energy Project - Phase 2 Forum
Thread: Smaller wu = No resets! |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 9
|
|
| Author |
|
|
Loui_h20
Cruncher Joined: Aug 12, 2013 Post Count: 25 Status: Offline |
From the 15th of December I've noticed that you've been sending out much, much smaller wu's. Wu's that on my i7 can be completed in around.... 2.5 to 3.5 hours on all 8 threads!!!
----------------------------------------I'll tell you now that after losing so many days worth of work due to a normal 11-12 hour wu finishing and qchem or boinc resetting all of the rest, due to no hart beat, it's so nice to be back up and running on all 8 threads on a i7 with no resets!!!!! Well, only one, as I think my pc was finishing the last of the 11 hour wu's. Is this size wu going to be the size of things to come? Can I suggest you guys stick with this size or maybe experiment and make them just a little bit bigger in order to reduce the load on your servers maybe. Yay! Crunching on all 8 threads and no resets !!!!!!!  [Edit 2 times, last edit by Loui_h20 at Dec 18, 2013 10:48:55 AM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hi Loui_h20,
The molecules we investigate have different sizes. Right now we are working on some small molecules. But there are some molecules much larger than we have ever worked on. Harvard runs molecules requiring up to 8 GB on its Linux cluster. Right now WCG only runs molecules that require less than 1GB, but we intend to run bigger molecules when we upgrade the program. Enjoy the tiny sizes while you can. Sorry I can't offer a permanent supply of midget molecules. Lawrence |
||
|
|
Loui_h20
Cruncher Joined: Aug 12, 2013 Post Count: 25 Status: Offline |
Damn it! :-). Knew it was to good to be true.......
----------------------------------------8GB!!!.. faint! Okay, when you upgrade the program please, oh please, oh please, incorporate a delayed ....file ziping time. I think that's what causes the wu to resets or at least temopreraly dump the results file until x minutes after completion before ziping. I wonder if a boinc config file can be written to do that?.... If you start sending out larger molecules I think the wu reset problem will only get worse on Windows systems... from what I've experience anyway. And see if you can crowbar using GPU's into the upgrade. I stopped every other project when I finally jumped on the CEP2 bandwagon! Got 3 GPU's sitting here doing f-all right now.... And I've got a feeling there is an awful lot of GPU power around the world waiting for something like CEP2 or 3 to take advantage of it. Thank you for taking your time to reply. L. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
The "much much smaller" actually started somewhere after the 8th [running all 16 jobs, no skipping after job 12], just took a while to get the bigger ones out the system. The 18 hours limit extension just announced is I think a precursor to what's coming... the extra heavies. As usual we'll approach those gradually, run times steadily increasing [if this follows the multi cycle pattern]. See http://bit.ly/WCGCE1 [Yes it looks chaotic for the last few months in the main graph, but the 3rd mini chart at bottom shows the historical behavior well... we're now entering like the 5th or 6 library workout [If I count right ;]
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
P.S. The "small is great" led to a new day record of > 60K results validated in a day. Regrettably, the small has the adverse that the number of gaps between one result and the next has increased... report 1, fetch 1 default. Each time a new one is fetched it has to wait it's turn in the client buffer. There's ways to speed that process up [e.g. app_config.xml <max_concurrent> tag], but the drive to push this feature wont come until the v7 client gets general acceptance and promotion at WCG. Maybe the wait is on that requested 'staggered start' feature
 |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
We are glad that you enjoy crunching on the smaller wus :). Right now, we have rather big ones in the feeder for a couple more weeks and then there should be a whole host of smaller ones again.
Best wishes, Your CEP team |
||
|
|
Speedy51
Veteran Cruncher New Zealand Joined: Nov 4, 2005 Post Count: 1290 Status: Offline Project Badges: 
|
A couple of days ago I had to tasks of run .89 of an hour unfortunately I do not have the task names. I hope some more shorter work units come along soon. I think we are expecting some.
----------------------------------------I am currently working on E219646_ 18_J.57.C52H28N2S2Se.00049852.0.set1d06_ 0-- currently on a job 2 I hope it will checkpoint in the next 3 hours and 40 minutes otherwise I will have to abandon this work unit since even with hibernation it still restarts from the latest checkpoint which is at 45 minutes. So I will lose around 12 hours of computing time if task does not checkpoint CPU time is currently at 10 hours 49 minutes  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hi Speedy51,
Ouch! While I try to keep 1 CEP2 unit running on my computer, I have been known to eliminate CEP2 I wanted to run some Beta units that required resetting for testing. Sometimes CEP2 is just too demanding. Lawrence |
||
|
|
Speedy51
Veteran Cruncher New Zealand Joined: Nov 4, 2005 Post Count: 1290 Status: Offline Project Badges:
|
Hi Speedy51, Ouch! While I try to keep 1 CEP2 unit running on my computer, I have been known to eliminate CEP2 I wanted to run some Beta units that required resetting for testing. Sometimes CEP2 is just too demanding. Lawrence It checkpoint did it 13 hours 38 minutes about quarter of an hour before I am due to turn my computer off so I am thrilled. Good luck with your bata testing |
||
|
|
|