| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: Help Stop TB
Thread: Limited workunits for Help Stop TB |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 752
|
|
| Author |
|
|
SekeRob
Master Cruncher Joined: Jan 7, 2013 Post Count: 2741 Status: Offline |
From 6:06 minutes after the whole and half hour they got piped at some points in time... got 26 T325 on the Octo at 10:53 hours TTC when they take 16.5+... the feeder is still mal-adapted to the real runtimes [so happens 6:06 after midnight and noon UTC is the start of the statistics run]. This load is high enough to now not having to switch profile setting back and forth... good for 2.25 days... enough auto-hitting the scheduler to keep them trickling in.
Today's projection is slightly up from yesterday... 6500 validations. Nine days and some plotted data at https://bit.ly/WCGHST1 |
||
|
|
Jack007
Master Cruncher CANADA Joined: Feb 25, 2005 Post Count: 1604 Status: Offline Project Badges: 
|
I wanna be ornry and just leave my comp on HSTB while I'm gone....
----------------------------------------but kinda defeats the purpose of computer running 24/7 so i'll load up on aids. maybe 3 days of aids, then ask for TB.... decent compromise.  |
||
|
|
imakuni
Advanced Cruncher Joined: Jun 11, 2009 Post Count: 103 Status: Offline Project Badges: 
|
Quick question: why isn't a maximum of 1 task per thread implemented for early projects? It would help prevent huge, unfinisheable caches (like rbotterb's) while allowing more people to help.
----------------------------------------It also comes with the bonus of diversifying the computer pool, which is good for both for early bug hunting and get more broad time estimates.  Want to have an image of yourself like this on? Check this thread: https://secure.worldcommunitygrid.org/forums/wcg/viewthread_thread,29840 [Edit 1 times, last edit by imakuni at Apr 3, 2016 5:08:26 AM] |
||
|
|
gb009761
Master Cruncher Scotland Joined: Apr 6, 2005 Post Count: 2986 Status: Offline Project Badges: 
|
imakuni, that's a very good question, and one (if I remember correctly) has been raised before. I can't remember what (even if) the response from the WCG techs was though - it being so long ago.
---------------------------------------- |
||
|
|
SekeRob
Master Cruncher Joined: Jan 7, 2013 Post Count: 2741 Status: Offline |
It's been proposed for FAHB [long running too], 1 or even 2 per core having inter-task result dependencies, and the 'we'll think about it'... some oiling needed. It was indicated that HST also has inter-task dependencies i.e. only when results/batches are back, can new work be generated using the output as base, so there is valid point of here too considering the N per core limit. Got silver worth completed, but half silver worth in PV state... no wingman. Considering that production has sunk to 6-6.3K validations/day when 9.6K are theoretically circulated [100 per 30 minutes in quorum 2], somewhere there's a serious traffic jam, on hosts for one.
|
||
|
|
rbotterb
Senior Cruncher United States Joined: Jul 21, 2005 Post Count: 401 Status: Offline Project Badges:
|
The only reason I'm up to my neck still in HST1 WUs is that when I pulled a bunch of them about 4 days back, I just happen to got them when my laptop was crunching them in about 5 hours each. If that was the case for all of them, I would actually have used them up by this weekend. It was when the new ones went to 24+ hrs each that things have gotten interesting. Making good progress crunching - got 20 WUs left with 4-5 days to go before deadline. With my quad core, I may manage to get almost all of them done by the deadline yet - hope is still there.
I would agree that things are definitely feast or famine on these WUs, but then that seems to be the case with many new projects. Being one of the little guys, I'm kind of used to sucking air looking for new WUs while the big whale crunchers are sucking up everything available for the first several months of a project. So the fact I happen to get myself up to my neck with WUs this past week means I'll actually get a Bronze badge during the first 30 days of a project and get a good ways towards Silver - now I just have to do my part and have my little old laptop crunch its butt off for the next week to get my backlog of work done ...  |
||
|
|
rbotterb
Senior Cruncher United States Joined: Jul 21, 2005 Post Count: 401 Status: Offline Project Badges:
|
I have several WUs in PV mode too. But then I suspect I'm now the one everyone is waiting on - It's pure heck being everyone's wingman .... Crunch, Crunch, Crunch my butt off ....
|
||
|
|
Eric_Kaiser
Veteran Cruncher Germany (Hessen) Joined: May 7, 2013 Post Count: 1047 Status: Offline Project Badges: 
|
I'm not calling myself a big cruncher. However I let my HST1 queue run dry and switch to another project that may end soon.
----------------------------------------That'll keep me busy a couple of month before I revert to HST1  |
||
|
|
SekeRob
Master Cruncher Joined: Jan 7, 2013 Post Count: 2741 Status: Offline |
Top batch now HST1_000464... got 5
The TTC projections are of the  department. Everything received today got recalced from 9:54 to 14:27. The 11:56 from previous days had gone back to 7:50 or so this morning and now jumped again to 11:40. All wrong, courtesy of the <do_not_use_dcf/> server instruction. Sadly so, because the original DCF mechanism would quickly adept to actual on-host long runtimes, and slowly decrease if shorts came by, making the buffer balancing much easier, where the shorts are in great minority (At least not seen a T000 arriving for at least 4 days). |
||
|
|
gb009761
Master Cruncher Scotland Joined: Apr 6, 2005 Post Count: 2986 Status: Offline Project Badges:
|
Yes, I'd seriously question the wisdom of not using the '<do_not_use_dcf/> server instruction' - as, not only did the HST WU's that I receive have ridiculously short estimates, this also crosses other projects (I've received a bunch of OET WU's that have an estimated time of 18 minutes - when in actual fact, they're taking just under 3 hrs. to crunch).
---------------------------------------- |
||
|
|
|