Advanced search

Message boards : News : Updates on Planck app and website

Author Message
Profile Marius
Project administrator
Project developer
Project scientist
Avatar
Send message
Joined: 29 Jun 15
Posts: 416
Credit: 4,276
RAC: 0
Message 20981 - Posted: 3 Apr 2016, 14:28:04 UTC

Thanks to you guys we've almost hit the 10,000 Planck parameter simulations we wanted to do. Read within what we've learned from these, where were going next, and about a few minor updates to C@H.

Profile Marius
Project administrator
Project developer
Project scientist
Avatar
Send message
Joined: 29 Jun 15
Posts: 416
Credit: 4,276
RAC: 0
Message 20982 - Posted: 3 Apr 2016, 14:28:10 UTC - in response to Message 20981.
Last modified: 3 Apr 2016, 22:26:07 UTC

As you'll see on the stats page, we're very close to finishing 10k sims. We've been looking through the results, writing up the paper, and are excited to share these with you when we're done. Based on what we've seen so far, we've learned there are some interesting ways in which we'd like to tweak the simulations and look further, so there is another batch of jobs which I will add shortly (this will be reflected on the stats page as well). It may be that we don't finish another 10k before submitting the paper, but I think we will have enough, and more jobs can be run during the referring process to sharpen things up. In any case, we will still give two week notice before picking the winners for being thanked in the paper. And thank you to everyone that has contributed thus far!

Along with this, there's a number of small updates to the website which I've just made,


  • You now need to fill in a reCAPTCHA to create an account on the website, sorry robots :(
  • The logout button which accidently dissapeared last week is now back
  • Uploading a profile picture was broken at some point, but is now fixed
  • You will now see the camb_boinc2docker and planck_param_sims apps identified by the plan class "docker_vbox64_mt". This is a change to the name only and does not require any action on your part. Its more to let everyone know these are both running on the same Docker / Virtualbox based technology, which was not obvious otherwise. This change was undone, thanks to pschoefer for pointing out to me this had an unintended consequence.


Profile Phil1966
Send message
Joined: 10 Jul 13
Posts: 26
Credit: 3,212,542
RAC: 5,283
Message 20984 - Posted: 3 Apr 2016, 15:40:38 UTC - in response to Message 20982.

Hello Marius,
Thank you for the update.
Best Regards,
Phil1966

Profile pschoefer
Send message
Joined: 26 Mar 08
Posts: 9
Credit: 1,946,775
RAC: 7,094
Message 20986 - Posted: 3 Apr 2016, 21:26:29 UTC - in response to Message 20982.
Last modified: 3 Apr 2016, 21:26:45 UTC


  • You will now see the camb_boinc2docker and planck_param_sims apps identified by the plan class "docker_vbox64_mt". This is a change to the name only and does not require any action on your part. Its more to let everyone know these are both running on the same Docker / Virtualbox based technology, which was not obvious otherwise.


No action required unless an app_config.xml like the one in the FAQ is used. ;)

Profile Marius
Project administrator
Project developer
Project scientist
Avatar
Send message
Joined: 29 Jun 15
Posts: 416
Credit: 4,276
RAC: 0
Message 20987 - Posted: 3 Apr 2016, 22:03:56 UTC - in response to Message 20986.
Last modified: 3 Apr 2016, 22:27:10 UTC


  • You will now see the camb_boinc2docker and planck_param_sims apps identified by the plan class "docker_vbox64_mt". This is a change to the name only and does not require any action on your part. Its more to let everyone know these are both running on the same Docker / Virtualbox based technology, which was not obvious otherwise.


No action required unless an app_config.xml like the one in the FAQ is used. ;)


Ahhh crap, that's right, I did not think of that. Sorry about that to everyone that's using that file and needs to go in and change it. Actually, I believe that just removing the line specifying the plan class is OK, checking now... It is not.

OK, I'm going to revert the plan class back to "vbox64_mt". Since it really was only superficially necessary, I'd rather not force a ton of people to re-edit their configuration files unless I have a good reason to. Should be done in a few minutes.

EDIT: OK, the apps are back to plan class "vbox64_mt". Thanks for pointing that out to me.

GGPA [E.R.]
Send message
Joined: 1 Apr 16
Posts: 1
Credit: 370,550
RAC: 0
Message 21007 - Posted: 11 Apr 2016, 6:55:35 UTC - in response to Message 20987.
Last modified: 11 Apr 2016, 6:56:03 UTC

Hi Marius,
about the challange we are doing on Plank,
I suggest you to insert into the paper the name of both teams that won the first 10.000 WU phase and the second one.
This would be correct, on my opinion, because you made two challange instead of one.
The same could be done also for the first 3 users
Best regards
GGPA

Profile 25000ghz [Lombardia]
Send message
Joined: 18 Aug 15
Posts: 5
Credit: 3,503,450
RAC: 0
Message 21008 - Posted: 11 Apr 2016, 7:21:39 UTC

Hi,
I have a dual cpu pc E5-2670v2 and wu "Plank" work with 40 cpus but they go all wrong. How can I fix?

Best regards,
Andrea

frederikhk
Send message
Joined: 9 Oct 15
Posts: 4
Credit: 70,156
RAC: 0
Message 21009 - Posted: 11 Apr 2016, 9:37:53 UTC - in response to Message 21008.

Hi Andrea. Try setting BOINC to use only 32 cores as it seems the error log indicates that it can only use 1 to 32 cores at a time. In your case set boinc to use 80% of CPU's.

Profile Marius
Project administrator
Project developer
Project scientist
Avatar
Send message
Joined: 29 Jun 15
Posts: 416
Credit: 4,276
RAC: 0
Message 21010 - Posted: 11 Apr 2016, 12:14:47 UTC - in response to Message 21007.

Hi Marius,
about the challange we are doing on Plank,
I suggest you to insert into the paper the name of both teams that won the first 10.000 WU phase and the second one.
This would be correct, on my opinion, because you made two challange instead of one.
The same could be done also for the first 3 users
Best regards
GGPA

Hi GGPA, I appreciate the suggestion, we considered it for a bit, but decided it against it as from the beginning we have said we would give two week notice when the contest is over, and it would be a bit unfair to pick any winner at all with out having done this.

I do hope no one feels like we're "stringing" them along with a never-ending contest, that's not at all the goal. Mostly its just not easy to estimate when the paper is done due to a number of things (some of which are out of our control). My best guess is that between writing, computing results, internal reviews, etc... this paper is a little under a month from submission, which means the two week notice will come fairly soon. It may be before this new 10k is not finished before submission, which as mentioned before is OK for the science.

Also so that you are aware, after the paper is submitted and the contest is officially over, while we are waiting for the referee report, we will likely still run some planck_param_sims as we can use results in the revised version.

Thanks everyone, let me know if you have further questions/concerns.

Profile Marius
Project administrator
Project developer
Project scientist
Avatar
Send message
Joined: 29 Jun 15
Posts: 416
Credit: 4,276
RAC: 0
Message 21011 - Posted: 11 Apr 2016, 12:57:30 UTC - in response to Message 21009.
Last modified: 11 Apr 2016, 12:57:48 UTC

Hi Andrea. Try setting BOINC to use only 32 cores as it seems the error log indicates that it can only use 1 to 32 cores at a time. In your case set boinc to use 80% of CPU's.


Thanks for bringing this to my attention again. Indeed, I think Virtualbox may have a hard maxiumum of 32 cores, whereas for now our jobs don't obey this (and fail when they try >32). I need to look into a bit more to verify but I think this is the case. For now, yes, try limiting the number of cores to <32 with either your global BOINC CPU usage or the method described here.

I will try to push an update shortly to limit our jobs to <32 CPUs per job globally so you don't have to by hand.

frederikhk
Send message
Joined: 9 Oct 15
Posts: 4
Credit: 70,156
RAC: 0
Message 21012 - Posted: 11 Apr 2016, 13:43:10 UTC - in response to Message 21011.

Great to hear! Appreciate the quick answers.

Profile 25000ghz [Lombardia]
Send message
Joined: 18 Aug 15
Posts: 5
Credit: 3,503,450
RAC: 0
Message 21015 - Posted: 11 Apr 2016, 16:12:21 UTC - in response to Message 21009.

Hi Frederikhk, thx for your answer. Now use 32cpu for wu but the wu not work more then to 0,100%

Please have a solution?

Regards

Profile Marius
Project administrator
Project developer
Project scientist
Avatar
Send message
Joined: 29 Jun 15
Posts: 416
Credit: 4,276
RAC: 0
Message 21016 - Posted: 11 Apr 2016, 16:17:41 UTC - in response to Message 21015.

Hi Frederikhk, thx for your answer. Now use 32cpu for wu but the wu not work more then to 0,100%

Please have a solution?

Regards

See here (this applies to the planck_param_sims jobs as well).

Profile 25000ghz [Lombardia]
Send message
Joined: 18 Aug 15
Posts: 5
Credit: 3,503,450
RAC: 0
Message 21017 - Posted: 11 Apr 2016, 18:15:30 UTC - in response to Message 21016.

See here (this applies to the planck_param_sims jobs as well).[/quote]


Thanks Marius, with your instructions now I understand. My problem is to have a very slow ADSL

Profile 25000ghz [Lombardia]
Send message
Joined: 18 Aug 15
Posts: 5
Credit: 3,503,450
RAC: 0
Message 21018 - Posted: 11 Apr 2016, 18:42:43 UTC - in response to Message 21017.

Now the wu are processed but the load of the cpu work is 22/23%. Incredible processing slowness :-(

Profile Marius
Project administrator
Project developer
Project scientist
Avatar
Send message
Joined: 29 Jun 15
Posts: 416
Credit: 4,276
RAC: 0
Message 21019 - Posted: 11 Apr 2016, 18:47:43 UTC - in response to Message 21018.

Now the wu are processed but the load of the cpu work is 22/23%. Incredible processing slowness :-(

Hm, this user was having perhaps a similar issue. For them lowering the CPUs/job to 8 seemed to work, does that work for you?

Also, how exactly did you lower the number of CPUs, globally for BOINC, or via the app_config.xml file? How exactly are you measuring CPU usage?

Profile Marius
Project administrator
Project developer
Project scientist
Avatar
Send message
Joined: 29 Jun 15
Posts: 416
Credit: 4,276
RAC: 0
Message 21023 - Posted: 11 Apr 2016, 19:47:53 UTC - in response to Message 21019.

[conversation with 25000ghz moved over to this thread]

Message boards : News : Updates on Planck app and website