Message boards : News : Status and Plans, Saturday 29th September, 2012
Message board moderation

To post messages, you must log in.

AuthorMessage
Eric Mcintosh
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 12 Jul 11
Posts: 857
Credit: 1,619,050
RAC: 0
Message 24875 - Posted: 29 Sep 2012, 11:20:02 UTC

All running very smoothly indeed. Just a problem with deadline scheduling which I hope we can discuss and resolve on Monday, especially with some feedback from the BOINC meeting in London.
Also some hiccups on the CERN AFS infrastructure.
I am now hoping to prioritise the writing of my paper on numeric results reproducibility but I am continuing to run work for the next weeks as described in my new thread "Work Unit Description"
in the Message Board "Number Crunching".
I am also pondering how to best handle "very long"
jobs bearing in mind your feedback.
And of course I shall try and keep you informed.

Thankyou for your continued support. Eric.


ID: 24875 · Report as offensive     Reply Quote
Toby Broom
Volunteer moderator

Send message
Joined: 27 Sep 08
Posts: 798
Credit: 644,718,949
RAC: 234,660
Message 24876 - Posted: 29 Sep 2012, 13:47:53 UTC

Would it be possiable to break the longer jobs up into smaller ones?

You could always use the CPDN model with trickles?

Maybe may the longer jobs GPU ones?
ID: 24876 · Report as offensive     Reply Quote
Eric Mcintosh
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 12 Jul 11
Posts: 857
Credit: 1,619,050
RAC: 0
Message 24877 - Posted: 29 Sep 2012, 19:12:29 UTC - in response to Message 24876.  

Indeed that is what I am thinking.
GPU's are on hold as we might get some help with that.
Eric.

ID: 24877 · Report as offensive     Reply Quote
Profile Tom95134

Send message
Joined: 4 May 07
Posts: 250
Credit: 826,541
RAC: 0
Message 24878 - Posted: 30 Sep 2012, 17:40:45 UTC

Regarding GPU Tasks, when it happens I would suggest that GPU Tasks be relatively short, e.g., one hour max and preferable 20~30 minutes in duration.

I run SETI@Home and Einsten@Home. However, I do not allow GPU Tasks for Einstein because then have, in the past, been very long and hog the GPU (since there is no provision for Task switching in the GPU).

Just my two cents worth.

Tom
ID: 24878 · Report as offensive     Reply Quote
William C Wilson
Avatar

Send message
Joined: 11 Sep 08
Posts: 25
Credit: 384,225
RAC: 0
Message 24880 - Posted: 1 Oct 2012, 13:32:03 UTC - in response to Message 24878.  

Tom, excellent point. I have been running SETI almost entirely with GPU (NVIDIA) and waiting with my arms crossed, tapping my feet, when LHC will start doing same. To use GPU, I would prefer task switching and settings to be available.
William C Wilson
São Paulo Brazil
ID: 24880 · Report as offensive     Reply Quote
Eric Mcintosh
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 12 Jul 11
Posts: 857
Credit: 1,619,050
RAC: 0
Message 24881 - Posted: 1 Oct 2012, 18:17:11 UTC

Well after some discussion today and your feedback I think I
can say we will split into units of not more than one hour.
This would be extremely satisfactory all round. The problem is
when can I do this? I think it is a top priority. It would also allow
us to run over one million turns rather easily. The resources
required here at CERN will be greater as we need to submit and retrieve
much more data. Nonetheless in theory we have a plan for more
here with lxtrack with more disk space and CPU.
In addition this could all be tested using our backup/test server to
avoid disruption to the current rather fine service.
I'll keep you posted. Eric.
ID: 24881 · Report as offensive     Reply Quote
Eric Mcintosh
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 12 Jul 11
Posts: 857
Credit: 1,619,050
RAC: 0
Message 24886 - Posted: 9 Oct 2012, 13:25:47 UTC - in response to Message 24885.  

Thanks for that; I would like to run just a few more monsters,
10**7 turns, o complete my "experiment". I could mix them with
the current scan. I need to see with Igor how that might be done.
Eric.
ID: 24886 · Report as offensive     Reply Quote
Snow Crash

Send message
Joined: 29 May 11
Posts: 10
Credit: 2,810,862
RAC: 0
Message 24888 - Posted: 9 Oct 2012, 23:54:53 UTC

I'm all for *monster* WUs ... bring them on I say!

While it might be a fair amount of work initially to set up a second application for people to choose from it would give you the flexibility to allow people to 'opt in' and also give you a seperate queue for any future "odd" runs you might dream up :-)

-Steve
ID: 24888 · Report as offensive     Reply Quote
Profile tazzduke

Send message
Joined: 24 Jun 10
Posts: 39
Credit: 4,971,347
RAC: 4,992
Message 24889 - Posted: 10 Oct 2012, 2:56:17 UTC - in response to Message 24888.  

I would love to get my hands on some of these monster wu's, am with Snow Crash on this one

Regards
Tazzduke
ID: 24889 · Report as offensive     Reply Quote
Profile BHoovsU
Avatar

Send message
Joined: 20 Feb 08
Posts: 4
Credit: 3,648,306
RAC: 0
Message 24891 - Posted: 14 Oct 2012, 18:45:54 UTC

I know this may be a little bit off topic, but I noticed that the boinc all project stats hasn't been updated for LHC in 302 days according to my profile page. It shows me at 7050 in credits when I actually have 46945 credits. Can you please check on that? Barry Hoover(BHoov)



http://allprojectstats.com/s299584a20a.png
ID: 24891 · Report as offensive     Reply Quote
Richard Haselgrove

Send message
Joined: 27 Oct 07
Posts: 186
Credit: 3,297,640
RAC: 0
Message 24893 - Posted: 15 Oct 2012, 8:48:14 UTC - in response to Message 24891.  

I know this may be a little bit off topic, but I noticed that the boinc all project stats hasn't been updated for LHC in 302 days according to my profile page. It shows me at 7050 in credits when I actually have 46945 credits. Can you please check on that? Barry Hoover(BHoov)



The statistics for the LHC project (like every BOINC project) are simply generated once, and placed in a public location for any and every statistics site to collect and process.

Other major players in the statistics game - BOINCstats and BOINC Combined Statistics - are up to date, so LHC is doing what it needs to do. I suspect that allprojectstats may not have updated their records when LHC changed location and URL last year. That's something you'll have to take up with them.
ID: 24893 · Report as offensive     Reply Quote

Message boards : News : Status and Plans, Saturday 29th September, 2012


©2024 CERN