News:

Members can see all forum boards and posts. Non members can only see a few boards.  If you have forgotten your password use this link to change it.
https://forum.boinc-australia.net/index.php?action=reminder

Main Menu

Project Update

Started by JugNut, April 01, 2015, 03:51:31 AM

Previous topic - Next topic

JugNut

Anyone crunching Universe should delete any WU's with a  "9" in the workunit name.
Or they will eventually be deleted for you :(

As is mentioned on the Universe's home page. http://universeathome.pl/universe/

Example WU: universe_xray_9_20000_300001-350000_304187_1



EDIT: And a "10" in the WU as well..  Not impressed but life goes on.

Example WU: universe_xray_10_20000_300001-350000_304187_1

Daniel

Universe@Home has recently added a new application to investigate the spin evolution of black holes.

A detailed description of the problem can be found here ...
http://universeathome.pl/universe/forum_thread.php?id=88

Dataman

For Universe@Home crunchers. Recent publication in Nature magazine.
http://arxiv.org/pdf/1602.04531v2.pdf

2016-07-12: Universe@Home, Nature article!


Results from the Universe@Home project were published in the most prominent
scientific journal, Nature! The article depicts a stellar evolutionary scenario
that leads to the formation of double black holes, which are probable
sources of gravitational waves similar to those observed recently.

It is a sequel to the article from year 2010, which predicted that the first
detection of gravitational waves will rather origin from a double black hole
merger than from a
double neutron star merger. The latter had been favoured by the majority of
scientists until recent discoveries.

The links to the article are provided below. Please notice especially the
acknowledgements in the Nature publication.

Nature article:
http://arxiv.org/pdf/1602.04531v2 (this article is identical to the Nature
publication, which does not have a free access:
http://www.nature.com/nature/journal/v534/n7608/full/nature18322.html)

The 2010 paper:
http://arxiv.org/pdf/1004.0386v2


chooka03

Dam you Xeon users!!  :compbash: (Calin & Michael)
How's a man with a poor old, 4 core CPU meant to stay ahead of these beasts  :cry

I need to win lotto.

chooka03

Lookout!  Jugnuts felt his leading position was under threat and has turned the turbo boosters on LOL  :cheer1:
:wink
:oz:

JugNut

#5
Quote from: chooka03 on March 25, 2017, 03:37:25 PM
Lookout!  Jugnuts felt his leading position was under threat and has turned the turbo boosters on LOL  :cheer1:
:wink
:oz:

LOL  vroom,  vroom  :greet  https://www.youtube.com/watch?v=qfAqtFuGjWM

Hi chooka,
Yea Universe doesn't grant as much credit as some of the others but it's still a quality project and that's what counts right? Although it is the best crediting project i've found for my raspberry Pi 3's. 

I originally bought the Pi's for POGS but to date I haven't found a POGS app that works with the Pi 3's, but Universe would have been a close second pick anyway so....  still happy.

I see you and d_worms and a few of the regulars have had a good crack at it lately as well.  :thumbsup:
https://boincstats.com/en/stats/158/user/list/0/0/21/#1

Teams in 25th place but should easily move up a couple of place's in the coming days. https://boincstats.com/en/stats/158/team/list/0/0#25
    :rocks

Crunch em if ya got em...  :crazy


EDIT: Up one place to 24th spot :)

kashi

Seems like a few members are having a nice run on Universe. Just checked and I haven't crunched any for a almost a year. Will give it another go after finishing some WCG badge hunting in a few days.

Think Linux may have been a bit faster than Windows but Universe applications have been updated since then. So will try some on both OSs and see for myself.

JugNut

#7
Yea i've been having problems with running a linux VM it does horrible things to the miners.  Make them completely unusable even after trying to disable USB in the VM it still the same.  I suppose the next step would be to try & disable USB within linux?  Any thoughts kashi?



PS: Picked up a matched pair of E5-2670's cheap :) now for the hard part, which board & RAM. :(
It may be a longggggg build,  but i'm in no hurry.  Had any idea's since your build?

Dataman

#8
Quote from: kashi on March 25, 2017, 08:53:29 PM
Seems like a few members are having a nice run on Universe. Just checked and I haven't crunched any for a almost a year. Will give it another go after finishing some WCG badge hunting in a few days.

Think Linux may have been a bit faster than Windows but Universe applications have been updated since then. So will try some on both OSs and see for myself.
I have been running Universe on two RPi's for about a month with no problems. ~6.5K/day and ~4.5 hours/wu.

EDIT: Ooops that is Collatz. Universe is 20 hrs/wu and ~4K/day on two RPi's.  :dance:


kashi

#9
Well I started some Universe@Home on dual box. Tasks run longer than I expected but complete and upload without problems.

However something rotten in Denmark with Universe@Home I fear. Have posted to the forum there so hopefully get a reply unlike the post last October about the same thing which went unanswered.

Have kept crunching it for now, if I don't get an answer I'll give it a miss. Would be a waste to continue using computer resources and electricity on a project where the research database continues to be compromised by people using a dodgy hacked application to cheat.

Edit: False alarm, quick response from admin explained it is caused by VM Linux hosts with too little shmem configured. I've seen task details that seem fishy like this before on other projects that actually were due to hacked applications. That's my excuse for my suspicion anyway. Nah, I'm just quick on the draw, "Whenever I see anonymous hosts I reach for my revolver."

Tally ho, onward and upward at Universe@Home. Will try it on Windows later today and see if it's any better than Linux. Linux is on a VirtualBox VM and sometimes VMs slow down on dual socket computers when more than 1 CPU's worth of cores are used in the VM. Really should dual boot on dual socket boxes rather than use VMs but prefer not to lose Samsung Magician Rapid Mode which reduces writes on my SSD.

JugNut

#10
Now i'm suspicious kashi...
You could very well be right mate, I clicked through hundreds of my work units and 98% of the wingmen were an anonymous host running Linux 3.2.0-56-virtual with either a  i3,  i5  or i7 and using one core in a VM.  As you mentioned they are usually quite fast too.  What would be the chances?  It would mean around 90 - 95% of all work is being done by people(or perhaps just one user) using a single core VM.  Again, what would be the chances?


And to make matters seem even more suss I just now noticed there has been a new app released for BHspin v2 on both windows & Linux.  The first change since July 2016.


PS: Most times theses things turn out to be nothing so i'll continue crunching Universe until there's clear proof one way or another.
PSS: Also the new version of BHspin v2 was supposed to fix a bug to do with long running tasks.

kashi

#11
Admin has posted that BHspin2 Version 0.02 has a bug so he's rolled back to Version 0.01. So I aborted any unstarted Version 0.02tasks  in my cache and just left a gaggle of the newly released UV tasks there. BTW UV application was also released today just to add to niggling synchronicity doubts, haha.

I'm running down the cache on the Linux VM and will switch to Windows. Received some reissued BHspin tasks that were the first I'd had without those anonymous wingmen with zero CPU time tasks and one of them was an E5-2670 same as mine also running Linux. His iteration of the work unit completed much quicker than mine so it's likely this VM is slow again because it has exceeded 16 cores. Either that or that wingman E5-2670 was running a smaller number of Universe tasks at once which can speed task processing up a lot.

Edit: Aha, the 0.02 version that was released to fix long running tasks caused that exact symptom for me. I left 17 Version 0.02 tasks running that had started and came back much later and they were at 6 hours, still not finished with percentage completed at between 96% and 97% and incrementing extremely slowly. So I aborted them, shut down Linux VM and started up BOINC on Windows. First task downloaded on Windows was 0.02 version, so I aborted it and reconfigured to ask for UV tasks only. You all know what's coming right? There were no UV tasks available. That's it, I'm out for now. Will do some Asteroids and check again next week.

kashi

Please excuse double post but I just found something potentially relating to these weird single core VM hosts that have been so prevalent on Universe@Home recently.

Hadn't noticed that these hosts were all single core until JugNut pointed it out.
I realised having multi core CPUs masquerading as multiple different single core hosts helps to hide high RAC and daily credit total if anything dodgy is happening.

However after today's schemozzle with BHspin2 Version 0.1 and Version 2.0, was looking around for another project to crunch. By chance happened upon a post in the DrugDiscovery@Home project with a download link to a virtual machine image required for an upcoming DrugDiscovery@Home application.

The admin at DrugDiscovery is also an admin at Universe. Everyone that has downloaded that VM image and uses it to run Universe@Home tasks will have the same BOINC version and Linux version. Maybe a long shot but it may explain some of the weirdness. Also those with little memory in their computers and/or no experience of correctly configuring VMs may cause those "Can't set up shared mem" error messages in Stderr output.

Just some possible explanations I thought I'd share. VM image is large so I haven't downloaded it to to see if it has the same BOINC version and Linux version as the mysterious Anonymous hosts. None of my current machines have much disk space as they both are running on 256 GB Samsung SSDs.

JugNut

#13
Mmm that's an interesting thought kashi I guess time will tell on that one. 
But now a new problem,  at the moment all my BHspin2 v0.02 tasks are just spinning there wheels and little else.  So far I haven't even completed even one of them.  They seem like they are behaving normally but get to about 94 - 98% they restart again at 0%.  I went and looked at the top users and no one seem to have completed any.  The only way you know this is happening is if you see it happen before your eyes.  The only other big clue is in the message logs I found all my PC were littered with this messages,  like...

Task universe_bh2_160803_104_41_20000_1-999999_280000_0 exited with zero status but no 'finished' file   
If this happens repeatedly you may need to reset the project.

And yes I did reset one PC but it will take hours to see if happens again.

I reported this on the forum

LOL Only days ago I was saying that this was a solid quality project,  boy how quickly things can change.  :pcwhack:
After reading a lot of older posts they seem to have had problems on & off since christmas :(

kashi

#14
Yes, I mentioned it in an edit to a previous post. Appears I should have made a separate new post about it. Then you may have seen it and been warned. My aversion to double posting has let you down. My apologies.

As mentioned in the actual post and not my later edit at the bottom, admin had already posted that Version 0.02 had a bug and he had to revert to Version 0.01.  But it seems he was unable to remove these faulty Version 0.02 tasks from the server and they continued to be given out as I also related in my edit. So thus my ditching the project completely and why I was looking for another project.