climateprediction.net home page
Deadlines and completion times don\'t make sense for sulphur cycle

Deadlines and completion times don\'t make sense for sulphur cycle

Questions and Answers : Windows : Deadlines and completion times don\'t make sense for sulphur cycle
Message board moderation

To post messages, you must log in.

AuthorMessage
nedsram-cdl

Send message
Joined: 14 Apr 05
Posts: 31
Credit: 16,491,691
RAC: 0
Message 16956 - Posted: 2 Nov 2005, 21:16:58 UTC
Last modified: 2 Nov 2005, 21:47:52 UTC

After updating BOINC to 5.2.5 (from 4.25) on three PCs, two have gone into \"panic mode\". Both are set up to run SETI and CPDN 50% each, but both refuse to run SETI at all. Both are running sulphur cycle. Having installed BoincView and looked at the deadlines, I can see why they are in panic mode, but what is being displayed seems utter rubbish (approximate values):

PC 1 (2 ghz pentium): CPU time 30 days, % done 43.7, to completion 90 days, deadline 26/1/06. At least the time to completion is going down steadily, but why is it so high?

PC 2 (750 mhz pentium): CPU time 19 days, % done 9.7, to completion 198 days, deadline 26/1/06. In this case the time to completion is going steadily up.

Apart from the shortness of the deadlines, the days to completion seems utter rubbish for PC 1, and the rate of change of days to completion seems utter rubbish for PC 2.

As things stand, it looks like PC 1 will get to run SETI in about a week, but PC 2 will not get to run SETI until hell freezes over. Of course I can override this manually, but in the long term it will just make matters worse.

The third PC I upgraded is also running CPDN and SETI, but it is running hadsm3, with a deadline of 28/8/06. This is not in panic mode.

It seems to me that there are problems with sulphur cycle: (1) the deadline is too close, (b) the completion time is being calculated incorrectly. Any comments please? In particular, is there any way of getting these PCs to report more sensible times?

Brian
ID: 16956 · Report as offensive     Reply Quote
Les Bayliss
Volunteer moderator

Send message
Joined: 5 Sep 04
Posts: 7629
Credit: 24,240,330
RAC: 0
Message 16959 - Posted: 2 Nov 2005, 22:15:32 UTC

Apparently the latest version of BOINC uses a \"learn mode\" to calulate completion times.
It\'s not very good at getting it right for cpdn, but it gets more accurate over time.
It has been stated elsewhere that it may take an entire sulphur model for BOINC to get it\'s figures right.

The sulphur deadline is short because the data is needed for the start of experiment 2 early next year.
However, it has also been stated that results from sulphur models that finish late will still be usefull.

You could try running cpdn on your fastest computer, and the other projects on the other computers.

ID: 16959 · Report as offensive     Reply Quote

Questions and Answers : Windows : Deadlines and completion times don\'t make sense for sulphur cycle

©2024 climateprediction.net