climateprediction.net home page
First Crack At CPDN/BOINC Stats

First Crack At CPDN/BOINC Stats

Message boards : Number crunching : First Crack At CPDN/BOINC Stats
Message board moderation

To post messages, you must log in.

AuthorMessage
old_user1
Avatar

Send message
Joined: 5 Aug 04
Posts: 907
Credit: 299,864
RAC: 0
Message 535 - Posted: 9 Aug 2004, 15:20:48 UTC

I\'ve had a first go at BOINC stats for CPDN, the logic is as follows:

1) trickles determine your \"credit\" until the end of a run. so some alpha testers that have brought their runs over are \"ahead\" since they may be in phase 2 or 3 already

2) the \"cobblestone equivalency\" seems to be right at 0.007 per timestep, so I\'ve put that in my master model table (i.e. different models in the future will have different complexities so may be more or perhaps less cobblestones per timestep)

3) I base it on model progress rather than just CPU time (although cpu time is tracked of course).

4) in result table: granted_credit should be your total credit per result to date (or per user, cpu, team etc), and claimed_credit would be the \"incremental\" credit (i.e. the difference from the last trickle I received)

5) in user/team/host tables I am just using my own \"validator\" based on the traditional BOINC one but that uses the \"incremental\" credit to tack onto the appropriate fields

6) hopefully this isn\'t too open to cheating and manipulation but we\'ll see


ID: 535 · Report as offensive     Reply Quote
Pconfig

Send message
Joined: 5 Aug 04
Posts: 84
Credit: 76,646
RAC: 0
Message 536 - Posted: 9 Aug 2004, 15:23:58 UTC

Looks good!
ID: 536 · Report as offensive     Reply Quote
Admiral_Hawkes

Send message
Joined: 5 Aug 04
Posts: 106
Credit: 1,886
RAC: 0
Message 543 - Posted: 9 Aug 2004, 15:40:23 UTC - in response to Message 535.  

> I've had a first go at BOINC stats for CPDN, the logic is as follows:
>
> 1) trickles determine your "credit" until the end of a run. so some alpha
> testers that have brought their runs over are "ahead" since they may be in
> phase 2 or 3 already
>
> 2) the "cobblestone equivalency" seems to be right at 0.007 per timestep, so
> I've put that in my master model table (i.e. different models in the future
> will have different complexities so may be more or perhaps less cobblestones
> per timestep)
>
> 3) I base it on model progress rather than just CPU time (although cpu time is
> tracked of course).
>
> 4) in result table: granted_credit should be your total credit per result to
> date (or per user, cpu, team etc), and claimed_credit would be the
> "incremental" credit (i.e. the difference from the last trickle I received)
>
> 5) in user/team/host tables I am just using my own "validator" based on the
> traditional BOINC one but that uses the "incremental" credit to tack onto the
> appropriate fields
>
> 6) hopefully this isn't too open to cheating and manipulation but we'll see
>
>
>
>

Yep great!!!

Founder of

<a href="http://climateapps2.oucs.ox.ac.uk/cpdnboinc/team_display.php?teamid=36"><img src="http://www.ug-abi.de/2008/projekt1.jpg"></a>
ID: 543 · Report as offensive     Reply Quote
Pconfig

Send message
Joined: 5 Aug 04
Posts: 84
Credit: 76,646
RAC: 0
Message 548 - Posted: 9 Aug 2004, 16:43:57 UTC
Last modified: 9 Aug 2004, 16:44:23 UTC

Carl, i found a little problem!

I was doing some rebooting while my pc was trickling and now the same trickle is reported and credited 3 times!

Time Sent (UTC) Phase Timestep CPU Time (sec) Average
09 Aug 2004 16:22:52 1 10802 25517 2.3622477319015
09 Aug 2004 16:19:35 1 10802 25517 2.3622477319015
09 Aug 2004 16:17:22 1 10802 25511 2.3616922792076

(dunno if it's credited three times you'll have to lookup yourself :D)
ID: 548 · Report as offensive     Reply Quote
old_user1
Avatar

Send message
Joined: 5 Aug 04
Posts: 907
Credit: 299,864
RAC: 0
Message 549 - Posted: 9 Aug 2004, 16:59:22 UTC - in response to Message 548.  

&gt; Carl, i found a little problem!
&gt;
&gt; I was doing some rebooting while my pc was trickling and now the same trickle
&gt; is reported and credited 3 times!
&gt;
&gt; Time Sent (UTC) Phase Timestep CPU Time (sec) Average
&gt; 09 Aug 2004 16:22:52 1 10802 25517 2.3622477319015
&gt; 09 Aug 2004 16:19:35 1 10802 25517 2.3622477319015
&gt; 09 Aug 2004 16:17:22 1 10802 25511 2.3616922792076
&gt;
&gt; (dunno if it's credited three times you'll have to lookup yourself :D)
&gt;

OK, I'll check tomorrow -- I think the way I did the stats it's OK since I check max phase/timestep so even if you came in with three 1/10802 trickles the "differential" is 0 added to the totals, but we'll see tomorrow.

ID: 549 · Report as offensive     Reply Quote
old_user412

Send message
Joined: 7 Aug 04
Posts: 18
Credit: 70,985
RAC: 0
Message 595 - Posted: 10 Aug 2004, 1:12:06 UTC - in response to Message 549.  
Last modified: 10 Aug 2004, 1:29:35 UTC

&gt; &gt; Carl, i found a little problem!
&gt; &gt;
&gt; &gt; I was doing some rebooting while my pc was trickling and now the same
&gt; trickle
&gt; &gt; is reported and credited 3 times!
&gt; &gt;
&gt; &gt; Time Sent (UTC) Phase Timestep CPU Time (sec) Average
&gt; &gt; 09 Aug 2004 16:22:52 1 10802 25517 2.3622477319015
&gt; &gt; 09 Aug 2004 16:19:35 1 10802 25517 2.3622477319015
&gt; &gt; 09 Aug 2004 16:17:22 1 10802 25511 2.3616922792076
&gt; &gt;
&gt; &gt; (dunno if it's credited three times you'll have to lookup yourself :D)
&gt; &gt;
&gt;
&gt; OK, I'll check tomorrow -- I think the way I did the stats it's OK since I
&gt; check max phase/timestep so even if you came in with three 1/10802 trickles
&gt; the "differential" is 0 added to the totals, but we'll see tomorrow.
&gt;
&gt;
&gt;
Hi Carl,
I can confirm that this is a problem with the stats! I had been running the model and was waiting for it to do a trickle at TS 21604. It had saved the model at TS 21600 (144 * 150) ** Why aren't trickles sychronised with 144 timestep boundaries when model is saved to hard drive!! ** After the model trickled, I checked the stats and my credit was ~ 150 ( the correct figure I think ). I then shut down the computer. A few hours later I restarted the computer and the model restarted at TS 21600. When it reached TS 21604 the model trickled again and when I checked my stats, the credit had jumped to ~ 226 from 150 in just 4 timesteps! When I start again in the morning I'll be up to ~ 300 credits. Easy way to get to the top of the leader board!!
Apart from this little buglet, the stats are looking good!
BTW would it be possible for you to set my credits to the correct figure - I don't want the same hassle as certain other people have had in the past!
ID: 595 · Report as offensive     Reply Quote
Profile astroWX
Volunteer moderator

Send message
Joined: 5 Aug 04
Posts: 1496
Credit: 95,522,203
RAC: 0
Message 599 - Posted: 10 Aug 2004, 2:15:19 UTC
Last modified: 10 Aug 2004, 2:58:11 UTC

Curious, following Carl's posts, I checked the leader boards. Through some fluke of fate, ...

Edit: (Censored, post hoc, by writer.)
________________________________________________
We have met the enemy and he is us -- Pogo
ID: 599 · Report as offensive     Reply Quote
old_user194

Send message
Joined: 5 Aug 04
Posts: 63
Credit: 21,399,117
RAC: 0
Message 604 - Posted: 10 Aug 2004, 5:16:57 UTC

s/ts to 13 decimal places - obviously still using some of the old S@H code. :)

Cheers,

PeterV.
ID: 604 · Report as offensive     Reply Quote

Message boards : Number crunching : First Crack At CPDN/BOINC Stats

©2024 cpdn.org