| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: OpenZika
Thread: Estimated runtime on ZIKA totally off |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 34
|
|
| Author |
|
|
TPCBF
Master Cruncher USA Joined: Jan 2, 2011 Post Count: 1930 Status: Offline Project Badges: 
|
Well, since I activated the ZIKA project for my hosts, I received about 180 across at least a dozen different hosts.
----------------------------------------One thing that I noticed that they all come in as "shorties", with initial runtime estimates in the 12-15min range. But that estimate seems to be totally off, the first 4 that I "fast tracked" on one of the hosts are now running for 60 minutes, with 33-52% done and remain time all showing in the 14-15 min range... Hope all is going well with those WUs now that the flood gates have been opened on that project...  Ralf  [Edit 1 times, last edit by TPCBF at May 21, 2016 5:40:43 PM] |
||
|
|
SekeRob
Master Cruncher Joined: Jan 7, 2013 Post Count: 2741 Status: Offline |
Yes, that's totally normal... follow the note by deltavee in the welcome thread how to force BOINC to compute TTC correctly for running tasks (
Ready to Start will always be off is the prediction, just like OET most always is... now don't spill your coffee again ;P) |
||
|
|
TPCBF
Master Cruncher USA Joined: Jan 2, 2011 Post Count: 1930 Status: Offline Project Badges:
|
Yes, that's totally normal... follow the note by deltavee in the welcome thread how to force BOINC to compute TTC correctly for running tasks ( Well, no it isn't. Ready to Start will always be off is the prediction, just like OET most always is... The estimates on OET or other projects might be off by some degree, but not by several 100% For example looking at my iMac that just got 14 ZIKA WUs, OET comes in with an estimate of 56min and finishes in 60-65. MCM comes in with 5h45min and finishes with 6h-6h30min. The only time that I have seen the estimates THAT far off was with UGM on a couple of hosts where those WUs would end up never finishing before the deadline and I aborted any new UGMs on those immediately. Other projects, like OET, MCM or FAH run on those hosts just fine. That's a huge difference to what I see at the moment with the ZIKA WUs and I am kind of afraid that there is the possibility that those end up just as the before mentioned UGM WUs. Hope that I am wrong though... now don't spill your coffee again ;P) Well, I will make sure to put down the mug and swallow the last gulp properly before reading posts in here for the time being...  Ralf  |
||
|
|
SekeRob
Master Cruncher Joined: Jan 7, 2013 Post Count: 2741 Status: Offline |
OK last attempt to get through: Some of the beta were short running, then a mechanism was developed to package multiple jobs in a single task to get an object runtime. If those jobs each have variable runtimes, and the mix of those is variable as well, it becomes potentially highly variable. If the generated work was headered with that beta23 feedback, the runtime will be initially very off. Once the production feedback loop gets real times, those go into new work, regardless if long or short, and the 6 Flags roller coaster continues to run. The <fraction_done_exact> tag helps to very quickly get adjusted TTCs, but only on started tasks. The fpops in the RtS rule, with WCG using <dont_use_dcf>. Nuff said on that topic.
----------------------------------------[Edit 2 times, last edit by SekeRob* at May 18, 2016 6:51:36 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
The 12 workunits from batches 00007 and 00012 that have completed on my PCs have taken between 1 and 3 hours, with between 2 and 65 tasks per workunit. Initial TTC for all of them was around 10 minutes.
|
||
|
|
SekeRob
Master Cruncher Joined: Jan 7, 2013 Post Count: 2741 Status: Offline |
See https://secure.worldcommunitygrid.org/forums/wcg/viewpostinthread?post=521980 for tech response.
|
||
|
|
pcwr
Ace Cruncher England Joined: Sep 17, 2005 Post Count: 10903 Status: Offline Project Badges: 
|
First 1 my laptop is running, got stuck at 0.6% for 8hrs, after a reboot, started again at 0.9% for 11mins.
----------------------------------------Patrick  |
||
|
|
Crystal Pellet
Veteran Cruncher Joined: May 21, 2008 Post Count: 1316 Status: Offline Project Badges: 
|
I returned the first 14 from a Linux VM on an Opteron 6172 out of 8 different batches (98-111). Inside jobs 18 up to 29.
----------------------------------------Average runtime 2 hours and 17 minutes (min 2 hrs; max 2.48 hrs) On the same host a Win VM is heading for 3-4 hours runtime.  |
||
|
|
SekeRob
Master Cruncher Joined: Jan 7, 2013 Post Count: 2741 Status: Offline |
Yes, being AD Vina science engine based, the Linux runtimes will be far far faster than on Windows. On first starting manually under Ubuntu 16.04 with 4.6 kernel, initially let it run without <fraction_done_exact/>, showing 19 minutes after 2 minutes, then applied the tag after 5 minutes and the TTC went to 4:38. Now at 1:30 it shows as 1:57 remaining, just not 3:30 to complete [at 1.6 Ghz]. As it is, the task header fpops will always reflect the project median runtime whence some production data has been collected... too high.
----------------------------------------[Edit 1 times, last edit by SekeRob* at May 19, 2016 7:47:00 AM] |
||
|
|
SekeRob
Master Cruncher Joined: Jan 7, 2013 Post Count: 2741 Status: Offline |
The learning has already progressed... when the first Zika arrived it showed 19 minutes TTC, just now got a second on the Linux and it shows 1:09 hours... the feedback loop has kicked in. The initial 6 Zika are still sitting on the Windows machine in queue at 17:22 minutes. A small hope that on the next receipt, the TTC will be retroactively adjusted... seen it happen on occasion :O)))
----------------------------------------[Edit 1 times, last edit by SekeRob* at May 19, 2016 9:01:09 AM] |
||
|
|
|