climateprediction.net (CPDN) home page
Thread 'Redundant Workunits'

Thread 'Redundant Workunits'

Message boards : Number crunching : Redundant Workunits
Message board moderation

To post messages, you must log in.

AuthorMessage
ChrisD

Send message
Joined: 8 Aug 04
Posts: 69
Credit: 1,561,341
RAC: 0
Message 39542 - Posted: 13 Apr 2010, 7:48:21 UTC

I have almost finished crunching this WU:

Name hadam3p_mya8_1960_2_006146266_2
Workunit 6412323

Created 7 Mar 2010 17:51:46 UTC
Sent 7 Apr 2010 19:18:17 UTC

Now checking out the WU

Workunit details
application UK Met Office HADAM3P
created 27 Apr 2009 18:24:47 UTC
name hadam3p_mya8_1960_2_006146266
minimum quorum 2

Computer# 874033 completed successfully at 17 Jun 2009 15:02:29 UTC
and computer#
826426 completed successfullly at 8 Jul 2009 22:53:29 UTC


So, 9 Months ago this WU was successfully processed by 2 computers. (minimum quorum filled)

Why do we have to crunch this one again?

There must be a good reason, I just can not find it, so could somebody enlighten me?

Tell me that I am wrong when I have the feeling that WU\'s are re-issued just to keep crunchers happy?

?? ChrisD






ID: 39542 · Report as offensive     Reply Quote
Les Bayliss
Volunteer moderator

Send message
Joined: 5 Sep 04
Posts: 7629
Credit: 24,240,330
RAC: 0
Message 39545 - Posted: 13 Apr 2010, 8:42:05 UTC

Have you read this post by Milo?
Or any of the dozens of replies through the years in answer to the same question?
(You\'ll need to use the Advanced part of the search function at the top of the page, and set it to 1 year.)
I posted a long reply last year some time, quoting the research paper on the matter, but I don\'t remember which forum.

ID: 39545 · Report as offensive     Reply Quote
Profilemo.v
Volunteer moderator
Avatar

Send message
Joined: 29 Sep 04
Posts: 2363
Credit: 14,611,758
RAC: 0
Message 39547 - Posted: 13 Apr 2010, 9:58:58 UTC

Hi Chris

Even in the same workunit and on the same platform the models generate slightly different results, I expect because they\'re so long. On different platforms the differences are greater and whether the processors are AMD or Intel also affects the results. Sometimes CPDN wants 6 or 8 results from the same workunit and gets them.

Several years ago three of the moderators sent each other the same task. They all crunched it and compared their graphs. Slight differences were visible even on the low-resolution graphs that we see.
Cpdn news
ID: 39547 · Report as offensive     Reply Quote
ChrisD

Send message
Joined: 8 Aug 04
Posts: 69
Credit: 1,561,341
RAC: 0
Message 39568 - Posted: 17 Apr 2010, 20:47:13 UTC - in response to Message 39547.  

Sorry for the late reply, but I have been down with the flu, and I was\'nt really up to participating.

Luckily my crunchers did not take notice :)

Back on-line.
I get Your point, and my first cruncher just finshed its fourth HADAM3 model.

Two other crunchers are half way through their 16 WU share of HADAM3 runs. So far one crash and 7 completed successfully.

I switched one of my chrunchers to HADSM3 and got 2 new tasks that have not been done before.

This just might be more like my cup of tea.

I have been here since the very start, and I still remember the thrill when Model Year was added to our tag by some gifted programmer (BBC experiment) and the 2012 bug that generated a lot of traffic in the forums. (All models crashed on one specific day and we all had to start over again on a modified model.)

443,390 HADSM3 runs has been completed so far.

Completing one more is not that thrilling...

Less than 1 in 10 users are active now.

Maybe we really need something new to get crunchers back.

ChrisD

ID: 39568 · Report as offensive     Reply Quote

Message boards : Number crunching : Redundant Workunits

©2024 cpdn.org