climateprediction.net home page
Formula For computing HADAM3P Credits

Formula For computing HADAM3P Credits

Message boards : Number crunching : Formula For computing HADAM3P Credits
Message board moderation

To post messages, you must log in.

AuthorMessage
Billy Ewell 1931

Send message
Joined: 14 Aug 06
Posts: 22
Credit: 4,834,184
RAC: 16,855
Message 36651 - Posted: 8 Apr 2009, 22:37:40 UTC

I noticed that the basis for awarding credits for 3P work units is obviously quite different from all the other CPDN projects I have run; as an item of interest, would you explain the formula as I cannot break-the-code.

By the way, my current inventory of 12 cpus is fully committed to CPDN in support of a most worthy scientific undertaking.

Bill
Austin, Texas
ID: 36651 · Report as offensive     Reply Quote
Les Bayliss
Volunteer moderator

Send message
Joined: 5 Sep 04
Posts: 7629
Credit: 24,240,330
RAC: 0
Message 36653 - Posted: 9 Apr 2009, 1:29:07 UTC

It\'s discussed here.


Backups: Here
ID: 36653 · Report as offensive     Reply Quote
Billy Ewell 1931

Send message
Joined: 14 Aug 06
Posts: 22
Credit: 4,834,184
RAC: 16,855
Message 36655 - Posted: 9 Apr 2009, 3:07:07 UTC - in response to Message 36651.  

I noticed that the basis for awarding credits for 3P work units is obviously quite different from all the other CPDN projects I have run; as an item of interest, would you explain the formula as I cannot break-the-code.

By the way, my current inventory of 12 cpus is fully committed to CPDN in support of a most worthy scientific undertaking.

Bill
Austin, Texas


Clarification

I did a poor job of phrasing my question: On ALL of the CPDN work units I previously processed before HADAM3P, the \"Granted Credit\" value was always contained within the Timestep numerical value. For instance, a Timestep numerical value of 51840 would result in a granted credit of 5184. For the HADAM3P units however, there is no correlation that I can determine between the timestep and granted credit; and therefore, I cannot determine how current the incremental granted credit for a work unit is in relation to the Timestep in the Trickles. I realize the question is probably of the silly category but I enjoy on a continuing basis the progress and more detailed reporting of all my CPDN work units.

Thanks for the reply.

Bill
ID: 36655 · Report as offensive     Reply Quote
Les Bayliss
Volunteer moderator

Send message
Joined: 5 Sep 04
Posts: 7629
Credit: 24,240,330
RAC: 0
Message 36656 - Posted: 9 Apr 2009, 4:34:15 UTC

Ah, right.
It\'s just coincidence that the models you\'re run have credit that\'s numerically the same as the number of trickles.
However, if you look at this \'slab\' model that you ran in 2007, you can see that it\'s different too.

The credit is based on the amount of work that a reference computer has to do to fully complete a model. This gets adjusted a bit during beta testing, as we find out how different types of computer handle a new model type.

There are 3 differences at present which affect the amount of work needed:
1) The number of phases for multi phase models; 3 for slab and 4 for mid-Holocene.
2) The type of ocean; fixed for slab (which is why it\'s called a \'slab\' model), and dynamic for the Coupled Ocean model.
3) If it\'s a single or multi resolution model. The model currently being tested in beta is multi-res, \"normal\" resolution for most of the planet, with a small area of hi-res in a \'certain place\'. There are 2 of these at present: the Pacific North West (of America), and Southern Africa.

There used to be a page at the Met Office web site about this model, but they\'ve changed their web site recently, and I can\'t find it now.


Backups: Here
ID: 36656 · Report as offensive     Reply Quote
wateroakley

Send message
Joined: 6 Aug 04
Posts: 185
Credit: 27,083,655
RAC: 6,161
Message 36661 - Posted: 9 Apr 2009, 12:06:13 UTC

The Met Office climate science projects page is here: projects
The PRECIS Regional Climate Modelling System page is here: PRECIS
ID: 36661 · Report as offensive     Reply Quote
Profile mo.v
Volunteer moderator
Avatar

Send message
Joined: 29 Sep 04
Posts: 2363
Credit: 14,611,758
RAC: 0
Message 36678 - Posted: 10 Apr 2009, 20:24:22 UTC
Last modified: 10 Apr 2009, 20:29:03 UTC

The only real \'formula\' for deciding how many credits a particular model type will be awarded is that Tolu considers their running speed on reference computers he has in Oxford. He also takes into account what the Beta project testers and moderators say about relative running speeds and the mods of course keep an eye on members\' comments on the forums. The ideal is to award similar credit for all model types per hour of processing. However, the same model type may perform better on Linux or Windows, and some Macs seem to perform extremely well in comparison.

So there\'s a general procedure but no mathematical formula as such.
Cpdn news
ID: 36678 · Report as offensive     Reply Quote
Billy Ewell 1931

Send message
Joined: 14 Aug 06
Posts: 22
Credit: 4,834,184
RAC: 16,855
Message 36694 - Posted: 11 Apr 2009, 20:26:13 UTC

Thanks to les, Hagar and Mo for all the assistance. The CPDN project is a pleasure to work with and provide support to.

After some experimentation, it finally occured to me that the incremental \"granted credits\" for the hadam3p work-unit tasks are virtually almost equal to 2.75% of the numerical value of the Timestep: Currently one of my running hadam3p tasks (6312013) has a granted credit of 554.4 at the timestep value of 20160. Not scientific but it lets me monitor trickle and credit progress.

And a further note: The Timestep/Credit correlation does not apply, I found out, when there is more than one phase to a work unit. I was wrong when I made a statement to the contrary in my original (corrected) posting.

Bill





ID: 36694 · Report as offensive     Reply Quote

Message boards : Number crunching : Formula For computing HADAM3P Credits

©2024 climateprediction.net