Message boards : News : Very long jobs
Message board moderation

To post messages, you must log in.

1 · 2 · Next

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 24688 - Posted: 22 Aug 2012, 15:48:40 UTC

I am now going to submit just a few hundred very
log 10**7 turn jobs to complete two studies.
I think this will be OK now; we shall see.
ID: 24688 · Report as offensive     Reply Quote
Ruud van der Kroef

Send message
Joined: 16 Aug 05
Posts: 5
Credit: 2,795,425
RAC: 0
Message 24689 - Posted: 22 Aug 2012, 17:16:07 UTC - in response to Message 24688.  

I think I already received one: http://lhcathomeclassic.cern.ch/sixtrack/workunit.php?wuid=2801939, task #6150030.
It is running 13h30m (9h50m effective) at 9.4%, and BoincTasks estimates that it will take another 6 days to finish.

Ruud
ID: 24689 · Report as offensive     Reply Quote
Profile Tom95134

Send message
Joined: 4 May 07
Posts: 250
Credit: 826,541
RAC: 0
Message 24690 - Posted: 22 Aug 2012, 19:51:18 UTC - in response to Message 24688.  

I am now going to submit just a few hundred very
log 10**7 turn jobs to complete two studies.
I think this will be OK now; we shall see.

I just looks at the "Server Status" and it shows 214,982 Tasks. Is this correct? That's a lot more than a "few hundred". Are these all test jobs?
ID: 24690 · Report as offensive     Reply Quote
Richard Haselgrove

Send message
Joined: 27 Oct 07
Posts: 186
Credit: 3,297,640
RAC: 0
Message 24691 - Posted: 22 Aug 2012, 20:16:35 UTC

I think I've got a very big, very long one: wlxu2_nuebb0__3__s__64.31_59.32__4_4.5__7__27_1_sixvf_boinc42970

Fortunately, it got sent to the very big, very fast computer I built at the weekend - an excellent test, thank you :)

It's reached 60% after 19 and a half hours - estimating another 14 to go. But the deadline is 28 August, like the others...

[and since receiving it I've received - and returned - other, shorter jobs. Not all the jobs in the queue are like this one]
ID: 24691 · Report as offensive     Reply Quote
Profile Magic Quantum Mechanic
Avatar

Send message
Joined: 24 Oct 04
Posts: 1114
Credit: 49,501,728
RAC: 4,157
Message 24692 - Posted: 22 Aug 2012, 20:26:34 UTC - in response to Message 24691.  



Fortunately, it got sent to the very big, very fast computer I built at the weekend - an excellent test, thank you :)



Nice one Richard,

I should have my GeForce 660Ti superclocked in a couple days......but just a quad-core running it and not close to your 8 cores.

I think you just might like your new machine.




Volunteer Mad Scientist For Life
ID: 24692 · Report as offensive     Reply Quote
Richard Haselgrove

Send message
Joined: 27 Oct 07
Posts: 186
Credit: 3,297,640
RAC: 0
Message 24693 - Posted: 22 Aug 2012, 20:34:32 UTC - in response to Message 24692.  

Fortunately, it got sent to the very big, very fast computer I built at the weekend - an excellent test, thank you :)


Nice one Richard,

I should have my GeForce 660Ti superclocked in a couple days......but just a quad-core running it and not close to your 8 cores.

I think you just might like your new machine.

It's still only 4 cores, but with the new i7 hyperthreading which seems to work much better than the P4 version did.

The supplier matched up the CPU/motherboard/RAM before I picked it up, so it ran at 4.5 GHz at first boot - and it's been running like that ever since.

Yes, I like it!
ID: 24693 · Report as offensive     Reply Quote
Jeroen

Send message
Joined: 25 Nov 05
Posts: 1
Credit: 969,430
RAC: 0
Message 24694 - Posted: 23 Aug 2012, 4:05:15 UTC - in response to Message 24688.  
Last modified: 23 Aug 2012, 4:11:08 UTC

I believe I have received three of the very long jobs so far. Two jobs have completed with the runtime of one job at 65,229 seconds and the second job at 94,123 seconds. The third job is presently at just over 95% with 27:57 runtime so far.

These jobs are indeed taking much longer compared to previously received jobs. I have my CPU running at half load with HT enabled. I think these jobs would take even longer if I were to load the CPU 100%.
ID: 24694 · Report as offensive     Reply Quote
Profile tazzduke

Send message
Joined: 24 Jun 10
Posts: 39
Credit: 4,971,347
RAC: 4,992
Message 24696 - Posted: 23 Aug 2012, 10:30:05 UTC - in response to Message 24694.  

Greetings

Have picked up 7 long ones and crunching at a good pace, glad to see the work has picked up

Regards
Tazzduke
ID: 24696 · Report as offensive     Reply Quote
Profile tazzduke

Send message
Joined: 24 Jun 10
Posts: 39
Credit: 4,971,347
RAC: 4,992
Message 24697 - Posted: 23 Aug 2012, 14:49:31 UTC

In further to my last post, I have set to NNT until I have crunched these 7 long workunits.

Regards
Tazzduke
ID: 24697 · 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 24698 - Posted: 24 Aug 2012, 6:07:15 UTC

Just to clarify most of the current jobs are part of the
intensity scan which I mentioned earlier.
The very very long jobs all begin wlxu7_*
ID: 24698 · Report as offensive     Reply Quote
Profile Conan
Avatar

Send message
Joined: 6 Jul 06
Posts: 108
Credit: 661,871
RAC: 196
Message 24699 - Posted: 24 Aug 2012, 6:32:49 UTC - in response to Message 24698.  

Just to clarify most of the current jobs are part of the
intensity scan which I mentioned earlier.
The very very long jobs all begin wlxu7_*


The "wlxu2_*" work unit type is the type that has been reported in this column and is the type that I have that report possible run times of 50 odd hours.

Could also include wlxu7_* but I have not seen any of these.

Conan
ID: 24699 · Report as offensive     Reply Quote
metalius
Avatar

Send message
Joined: 3 Oct 06
Posts: 101
Credit: 8,985,206
RAC: 43
Message 24700 - Posted: 24 Aug 2012, 6:36:51 UTC - in response to Message 24688.  
Last modified: 24 Aug 2012, 6:41:13 UTC

I am now going to submit just a few hundred very
log 10**7 turn jobs to complete two studies.
I think this will be OK now; we shall see.

No, Eric!
This is not Ok, because these very long tasks still have the same deadlines as normal tasks.
It is very sad, but You are ignoring some facts...
1. Almost nobody of us is crunching LHC@home only. So, such big tasks with extremely short deadlines are starting to run in panic mode immediately after they are downloaded. CPU tasks from another projects are blocked.
2. Not all of us are rich people from rich countries, so not all of us:
- have newest and fastest machines;
- have possibility to crunch 24/7 (if computer is at work, very often it is running only 8/5).
PLEASE, do not follow some arrogant projects, designed for US, British Union and EU Western only.
ID: 24700 · Report as offensive     Reply Quote
Richard Haselgrove

Send message
Joined: 27 Oct 07
Posts: 186
Credit: 3,297,640
RAC: 0
Message 24701 - Posted: 24 Aug 2012, 7:17:08 UTC - in response to Message 24699.  

Just to clarify most of the current jobs are part of the
intensity scan which I mentioned earlier.
The very very long jobs all begin wlxu7_*

The "wlxu2_*" work unit type is the type that has been reported in this column and is the type that I have that report possible run times of 50 odd hours.

Could also include wlxu7_* but I have not seen any of these.

Conan

The first batch of very long jobs we ran at the beginning of this month were wlxu7_*

The ones I'm working on now are all wlxu2_*
ID: 24701 · Report as offensive     Reply Quote
Uffe F

Send message
Joined: 9 Jan 08
Posts: 66
Credit: 727,923
RAC: 0
Message 24702 - Posted: 24 Aug 2012, 9:33:20 UTC

Could you please make the deadline for the long runs 2 weeks instead? I had to cancel 2 jobs because I couldn't do them in time (I ofcourse aborted them before starting them to not waste power)
ID: 24702 · 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 24703 - Posted: 24 Aug 2012, 10:28:36 UTC - in response to Message 24702.  

I need to see Igor about this...........
I set dealine of 30 days always. However I
think BOINC re-computes based on the estimated
FLOPS. Eric.
ID: 24703 · 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 24704 - Posted: 24 Aug 2012, 10:30:02 UTC - in response to Message 24701.  

There are ony afew hundred wlxu7. Ther arevvery
very many "long" tasks including wlxu2 but they
run for "only" a maximum of 10**6 i.e. one million
turns.
ID: 24704 · Report as offensive     Reply Quote
Profile Conan
Avatar

Send message
Joined: 6 Jul 06
Posts: 108
Credit: 661,871
RAC: 196
Message 24705 - Posted: 24 Aug 2012, 11:54:29 UTC - in response to Message 24704.  
Last modified: 24 Aug 2012, 11:55:33 UTC

There are only a few hundred wlxu7. There are very,
very many "long" tasks including wlxu2 but they
run for "only" a maximum of 10**6 i.e. one million
turns.


Ah, OK,
No problem, my 50 hour work units are not the real long ones. The wlxu7_ work units would be impressive then, 100 hours or more.

Conan
ID: 24705 · 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 24706 - Posted: 24 Aug 2012, 12:10:58 UTC

Now with this magnificent system (thanks to all) I am thinking
of 100 million turns next to study the growth of floating-point
error and the loss of symplecticity! We shall see.
ID: 24706 · 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 24707 - Posted: 24 Aug 2012, 12:18:27 UTC - in response to Message 24700.  

Du calme as we say in French. I am hurt by your accusation. I have always
insisted that we run on any "old" machine, even a Pentium 3. The experiment
with 10 million turns has shown some deadline/scheduling problems.
These will be addressed before we proceed. The few hundred jobs
are necessary to complete a couple of studies. I have a few hundred
thousand other jobs available.

Watch this space. Eric.
ID: 24707 · Report as offensive     Reply Quote
Profile jujube

Send message
Joined: 25 Jan 11
Posts: 179
Credit: 83,858
RAC: 0
Message 24710 - Posted: 24 Aug 2012, 20:06:44 UTC - in response to Message 24707.  

Du calme as we say in French. I am hurt by your accusation. I have always
insisted that we run on any "old" machine, even a Pentium 3. The experiment
with 10 million turns has shown some deadline/scheduling problems.
These will be addressed before we proceed. The few hundred jobs
are necessary to complete a couple of studies. I have a few hundred
thousand other jobs available.

Watch this space. Eric.


Tasks that are much longer than the tasks a project usually issues cause problems on some hosts. Also, it's entirely possible that your longest tasks won't finish before deadline on a Pentium 3 even if it crunches 24/7. Other projects have run into exactly the same problem and rather than keep all tasks short they have configured their server to offer 2 different types of tasks: long tasks and short tasks. They create an option in user website prefs to allow crunchers to choose both types or just one type depending on their hardware and expectations. I have no idea how much work it would take to implement such an option (perhaps server code already has that functionality?) but you/Igor might give it some thought. GPUgrid offers a long vs. short option, maybe they can give you/Igor some ideas on how to implement it.
ID: 24710 · Report as offensive     Reply Quote
1 · 2 · Next

Message boards : News : Very long jobs


©2024 CERN