Load test time
log in

Advanced search

Message boards : News : Load test time

Author Message
Profile Kevin
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist
Avatar
Send message
Joined: 27 Jul 12
Posts: 507
Credit: 14,344,251
RAC: 1,489
Message 450 - Posted: 29 Aug 2012, 6:58:30 UTC

Right folks - load test time.

David has been busy at Johns Hopkins University and made some modifications to his code for making the POGS image files, allowing substantially more pixels to be considered for a fit (now each band is masked independently). In this iteration (v3) he has used constant surface brightness limits across the sample with the exact value dependent upon band (28.0, 28.0, 26.0, 27.2, 27.1, 26.0, 25.0, 22.6, 21.0, 18.0, 16.2 ABmag/sq.arcsec for FUV, NUV, u, g, r, i, z, w1, w2, w3, w4).

We've released 188 galaxies for this load test.

Hopefully on Friday we'll have the "Lite" version ready for people to process 5 - 10 pixels a go on the more venerable systems out there.

I've also added limits to try and stop people sucking too many units
____________
Regards
Kevin
-----
International Centre for Radio Astronomy Research

Profile DoctorNow
Avatar
Send message
Joined: 4 Aug 12
Posts: 5
Credit: 106,774
RAC: 0
Message 452 - Posted: 29 Aug 2012, 9:32:22 UTC - in response to Message 450.

I've also added limits to try and stop people sucking too many units

How high did you set that?
I recognize my comp at home has got 29 WUs a while ago. Still a bit too much - at least for me.
____________
Life is Science, and Science rules. To the universe and beyond
Proud member of BOINC@Heidelberg
My BOINC-Stats

Profile Kevin
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist
Avatar
Send message
Joined: 27 Jul 12
Posts: 507
Credit: 14,344,251
RAC: 1,489
Message 453 - Posted: 29 Aug 2012, 9:39:12 UTC - in response to Message 452.

10 per Core
____________
Regards
Kevin
-----
International Centre for Radio Astronomy Research

Profile LawryB
Send message
Joined: 10 Aug 12
Posts: 16
Credit: 12,159,086
RAC: 0
Message 454 - Posted: 29 Aug 2012, 9:51:07 UTC - in response to Message 453.


Thanks for that Kevin.

Profile Mankka*
Avatar
Send message
Joined: 3 Aug 12
Posts: 39
Credit: 7,185,304
RAC: 0
Message 455 - Posted: 29 Aug 2012, 10:46:09 UTC

I think the 10 tasks/Core limit is a reasonable value to get the work done before the deadline.

I can also report that the limit seems to work fine, I only got 10 tasks/core on the empty boxes (ok, I have some "overloaded" boxes still crunching older wu's, but they will be done by deadline).

Nice work Kevin & with the amount of work you released, nobody shoud have any
problems to get something to crunch on =)

Profile DoctorNow
Avatar
Send message
Joined: 4 Aug 12
Posts: 5
Credit: 106,774
RAC: 0
Message 456 - Posted: 29 Aug 2012, 12:07:00 UTC - in response to Message 453.
Last modified: 29 Aug 2012, 12:07:18 UTC

10 per Core

Ok.
Well, to me 4 or 5 per core would sound more convenient - would fit better with the other work my comps are crunching on.
Does this also apply to when you deploy the lite version? I guess these WUs would be crunched a lot faster so probably the 10 per core are suitable for this.
____________
Life is Science, and Science rules. To the universe and beyond
Proud member of BOINC@Heidelberg
My BOINC-Stats

Profile Dingo
Avatar
Send message
Joined: 3 Aug 12
Posts: 120
Credit: 22,808,806
RAC: 32
Message 457 - Posted: 29 Aug 2012, 13:25:49 UTC - in response to Message 456.
Last modified: 29 Aug 2012, 13:26:42 UTC

It still depends on your setting in BOINC.

I only downloaded three work units as I have three cores available to BOINC and in the preferences "Minimum Work Buffer" is set to .50 days worth of work and "Maximum" is set to 0.00. Depending on your setting you will get 10 or any number under that.
____________

Proud Founder and member of



Have a look at my WebCam

Dataman
Avatar
Send message
Joined: 10 Aug 12
Posts: 12
Credit: 13,497,402
RAC: 25,880
Message 458 - Posted: 29 Aug 2012, 15:01:36 UTC

So far everything looks good to me. 10/core or less is fine with me. I have no errors but one invalid that I will look into later. Good to see new work. Thanks.
____________

Profile RAMen
Avatar
Send message
Joined: 9 Aug 12
Posts: 17
Credit: 4,068,701
RAC: 0
Message 459 - Posted: 29 Aug 2012, 15:30:57 UTC - in response to Message 450.
Last modified: 29 Aug 2012, 15:36:49 UTC

I am also fine with the cache limit of 10 per core.
Unfortunately I was filling the caches with new work units from other projects as the new galaxies were being uploaded so will be a while before i can get back to full production at POGS.
currenntly:
693 valid WU
0 invalid WU
2 errors - both killed from the server
So few errors from a new project in development is exceptional. Well done !!
____________

Profile Jose [Team Musketeers]
Avatar
Send message
Joined: 12 Aug 12
Posts: 4
Credit: 278,046
RAC: 0
Message 461 - Posted: 29 Aug 2012, 22:17:00 UTC

Excellent for me, great job Kevin!
I hope the new job, I have still some units on two machines.

Message boards : News : Load test time


Main page · Your account · Message boards


Copyright © 2017 The International Centre for Radio Astronomy Research