Am I going mad?

Section supporting all the various Climate Change projects: CPDN, CPDN Beta, Quake Catcher.
Post Reply
Rockinfroggi
Posts: 1434
Joined: Tue Jan 09, 2007 12:00 am

Am I going mad?

Post by Rockinfroggi »

I have 2 CPDN WUs on the go, which from time to time I suspend.
I've just noticed with one of them, even though it is suspended the CPU time is going up and the To Completion time is increasing at the same rate?

Is this a known phenomena or am I going madder than a hatter?



Gary.
Rockinfroggi
Posts: 1434
Joined: Tue Jan 09, 2007 12:00 am

Post by Rockinfroggi »

Hello Mr Doug 'u' Louse aka CPDN Cheerleader, is there anybody homeeeeeeeee



Annoying little Frog.
DJH@GB-Ro
Posts: 525
Joined: Wed Mar 15, 2006 12:00 am

Post by DJH@GB-Ro »

Rockinfroggi wrote:Hello Mr Doug 'u' Louse aka CPDN Cheerleader, is there anybody homeeeeeeeee



Annoying little Frog.
I'm always checking the boards and I haven't come accross this on CPDN :? I know it happens on Xtremlab. Have you resumed and re-suspended? Or restarted the machine/BOINC?
Rockinfroggi
Posts: 1434
Joined: Tue Jan 09, 2007 12:00 am

Post by Rockinfroggi »

I resumed it earlier today and it still does the same while it is waiting for it's turn to crunch.

The other WU on my laptop is fine and doesn't do this, so I'll try shutting Boinc down now and see if that stops it short term, though I know it will start again.

Thanks for the reply and I'll post if I notice anything else.


Gary.
UBT-mark3346
Active UBT Contributor 15+ yrs
Posts: 7515
Joined: Wed May 03, 2006 1:00 am
Contact:

Post by UBT-mark3346 »

You probably are aware anyway but make sure you have lots of cooling for your laptop if you are running CPDN on it.  :lol:
Image
Rockinfroggi
Posts: 1434
Joined: Tue Jan 09, 2007 12:00 am

Post by Rockinfroggi »

UBT-mark3346 wrote:You probably are aware anyway but make sure you have lots of cooling for your laptop if you are running CPDN on it.  :lol:

Thanks for that Mark, yes I have it sitting on a 2 fan laptop cooler but now the hot weather is here will be keeping a close eye on it's temp.



Gary.
UBT-mark3346
Active UBT Contributor 15+ yrs
Posts: 7515
Joined: Wed May 03, 2006 1:00 am
Contact:

Post by UBT-mark3346 »

Rockinfroggi wrote:.... yes I have it sitting on a 2 fan laptop cooler but now the hot weather is here will be keeping a close eye on it's temp.



Gary.
Laptop cooler..find out something new everyday......got myself one this lunchtime, try it when I get home, maybe I won't be able to use the laptop to fry eggs anymore :lol:
Image
Rockinfroggi
Posts: 1434
Joined: Tue Jan 09, 2007 12:00 am

Post by Rockinfroggi »

UBT-mark3346 wrote:
Rockinfroggi wrote:.... yes I have it sitting on a 2 fan laptop cooler but now the hot weather is here will be keeping a close eye on it's temp.



Gary.
Laptop cooler..find out something new everyday......got myself one this lunchtime, try it when I get home, maybe I won't be able to use the laptop to fry eggs anymore :lol:

They do help quite a bit, but come summer I usually direct a small desktop fan on both my laptops as well, just for good measure.

Just one little tip, most of the coolers I've seen have a lip at the front so you place the front of the laptop against this, I find it best to check the underside of the laptop and place the hottest area directly above the fans for maximum effect.  
Gary.
smith639
Posts: 94
Joined: Thu Apr 26, 2007 1:00 am

Post by smith639 »

I too am running a CPDN WU, obviously a long process (no pun intended).
Based on the time run / %ge complete, I estimate it will run for nearly 5000 hours rather than the 2150 hours it stated.
Look closely at mine too and the two 'timers' are both increasing, not always at a uniform rate.
To sum it up I am not certain that this will complete before the deadline of May 2008, even running 24/7
UBT-mark3346
Active UBT Contributor 15+ yrs
Posts: 7515
Joined: Wed May 03, 2006 1:00 am
Contact:

Post by UBT-mark3346 »

They are big ones thats for sure...Mine took a little while to settle down with the completion time, if you have completed more than 10% then working it out the way you have should be fairly accurate but even if it does go the full 5000 should still finish around the end of the year....running 24/7 with no other projects  :!:
Image
MikeMarsUK
BOINCSynergy team member
Posts: 110
Joined: Sat Aug 12, 2006 1:00 am

Post by MikeMarsUK »

The deadlines are ignored as far as CPDN itself goes (but your Boinc manager will go bananas with scheduling for other projects due to EDF etc).

The initial estimates are based on the Boinc benchmarks and hence are rubbish, as suggested calculating it yourself is much more accurate.  The important figure is the 'seconds per timestep'.  Multiply this by 4 million-and-a-bit.
Rockinfroggi
Posts: 1434
Joined: Tue Jan 09, 2007 12:00 am

Post by Rockinfroggi »

MikeMarsUK wrote:The deadlines are ignored as far as CPDN itself goes (but your Boinc manager will go bananas with scheduling for other projects due to EDF etc).

The initial estimates are based on the Boinc benchmarks and hence are rubbish, as suggested calculating it yourself is much more accurate.  The important figure is the 'seconds per timestep'.  Multiply this by 4 million-and-a-bit.
Any chance of pointing me in the direction of where this is explained as I've never looked that deeply into this project and would be nice to try and work it out.


Gary.
MikeMarsUK
BOINCSynergy team member
Posts: 110
Joined: Sat Aug 12, 2006 1:00 am

Post by MikeMarsUK »

There are a bunch of README postings here:

http://www.climateprediction.net/board/ ... m.php?f=36

But as far as I can remember they don't include anything about estimating the speed of the model.   So ...

* Each model is 160 years

* Each year consists of 360 days (12 months of 30 days).  At the end of each year (approx 4th Dec each model year), the climate data is extracted, compressed, and uploaded to the servers in the form of a 'trickle' (so it isn't a disaster if the model crashes before the end).

* There are 72 timesteps per model day (hence 25920 timesteps per year, 4,147,200 over the life of the model).  Note that 259.2 credit per yearly trickle = 0.01 credit per timestep.

To see the current s/ts figure (averaged over the life of the model), show the graphics, press 'Z' to hide the grey sidebar.  The s/ts speed should be on the lower-left.  Alternatively, you can follow the 'results' link, and you'll see a list of trickles and the average speed against each one.

Multiply this s/ts figure by 4,147,200 to get the total time in seconds of the model.

The quickest (but less accurate) way to estimate speed is to see how many hours each trickle (year) takes, and multiply that by 160.

The final way of calculating total time is to look at the cpu-hours, and divide by %done.  So %done = 25%, CPU hours = 1000, hence 1000 / 0.25 = 4000.
Rockinfroggi
Posts: 1434
Joined: Tue Jan 09, 2007 12:00 am

Post by Rockinfroggi »

Thanks for that Mark, I'll let my first coffee start my brain ticking then get calculating.



Gary.
Post Reply