climateprediction.net home page
BOINC user preferences

BOINC user preferences

Message boards : Number crunching : BOINC user preferences
Message board moderation

To post messages, you must log in.

AuthorMessage
ML1

Send message
Joined: 5 Aug 04
Posts: 21
Credit: 9,084,238
RAC: 0
Message 64 - Posted: 5 Aug 2004, 17:44:30 UTC

\" These apply to all BOINC projects in which you participate.
On computers attached to multiple projects, the most recently modified preferences will be used. \"


I\'m running both s@h and CP beta with a 100:950 split... (:-))

CP BOINC v4.02 shows slightly different default prefs to those for s@h. Also, in those defaults are the cache limits which will apply to \'all projects\'.

CPDN & s@h have very different WU timescales. I certainly don\'t want to hog 60 day\'s worth of s@h WUs!


Can BOINC be updated so that the cache defaults are overridden in the project specific prefs?

Or how will CP react if I knock the cache back down to the s@h defaults of 0.1 to 1 day??

Regards,
Martin

ID: 64 · Report as offensive     Reply Quote
old_user59

Send message
Joined: 5 Aug 04
Posts: 12
Credit: 115,984
RAC: 0
Message 69 - Posted: 5 Aug 2004, 17:58:58 UTC

I did not think you could use Boinc 4.xx for seti yet??? On the other hand I know your problem will happen when we can use them all.
ID: 69 · Report as offensive     Reply Quote
old_user73

Send message
Joined: 5 Aug 04
Posts: 39
Credit: 14,887
RAC: 0
Message 70 - Posted: 5 Aug 2004, 18:34:15 UTC - in response to Message 64.  

As far as I know it will work just right even if you just request 1 second of work (ie. 0.00001 days or something). So the default settings of climateprediction are way too high and will probably cause problems for other projects.
ID: 70 · Report as offensive     Reply Quote
old_user1
Avatar

Send message
Joined: 5 Aug 04
Posts: 907
Credit: 299,864
RAC: 0
Message 89 - Posted: 5 Aug 2004, 19:16:45 UTC - in response to Message 64.  

Hmmm, OK, I was trying to get more "CPDN-ish" timescales but I will double-check if that will cause trouble for other projects. Basically the problem is whether a request coming in will get sent at least one workunit as long as the user has enough disk space allocated for CPDN/BOINC etc.
ID: 89 · Report as offensive     Reply Quote
John McLeod VII
Avatar

Send message
Joined: 5 Aug 04
Posts: 172
Credit: 4,023,611
RAC: 0
Message 127 - Posted: 6 Aug 2004, 1:11:53 UTC - in response to Message 89.  

> Hmmm, OK, I was trying to get more "CPDN-ish" timescales but I will
> double-check if that will cause trouble for other projects. Basically the
> problem is whether a request coming in will get sent at least one workunit as
> long as the user has enough disk space allocated for CPDN/BOINC etc.
>
BOINC will download a WU from any project even if it asks for 1 second of work, and the WU is scheduled for much longer than that. This is the technique used if 3.x to get a WU for a multi-processor that is a WU short for it's CPU count. I have seen this many times as I have a couple of slow Dual CPU machines (Probably not suitable for CPDN as they would take about 1.5 years to finish a model.) There was also a bug in 3.x that caused this to happen at times that it shouldn't, overflowing the cache and sometimes the deadlines.

The deadline for S@H is 14 days, and the deadline for Predictor can be as short as a day. Folding@Home is rumored to be looking at a 5 day deadline.

Since there is only one cache for all projects, a reasonable number should be specified as a default. Any WU that is not processed within the deadline will likely be discarded, and granted no credit. A shorter default would be useful.

jm7
ID: 127 · Report as offensive     Reply Quote
Jord
Avatar

Send message
Joined: 5 Aug 04
Posts: 250
Credit: 93,274
RAC: 0
Message 172 - Posted: 6 Aug 2004, 11:33:22 UTC

Also, a report deadline of 11 months, isn't that a bit too much? Or are you waiting to see how quickly some machines will return the unit, before you scale the report deadline down, to allow for lesser machines to cope with it as well?

As for during the crunch reporting/claiming credits, CPDN Classic would do that. Is there an option for BOINC to report it similarly?
--------------------
Jordâ„¢
ID: 172 · Report as offensive     Reply Quote
Profile Keck_Komputers
Avatar

Send message
Joined: 5 Aug 04
Posts: 426
Credit: 2,426,069
RAC: 0
Message 177 - Posted: 6 Aug 2004, 12:16:16 UTC - in response to Message 172.  

> Also, a report deadline of 11 months, isn't that a bit too much? Or are you
> waiting to see how quickly some machines will return the unit, before you
> scale the report deadline down, to allow for lesser machines to cope with it
> as well?
>
> As for during the crunch reporting/claiming credits, CPDN Classic would do
> that. Is there an option for BOINC to report it similarly?
> --------------------
> Jordâ„¢
>
I have been following the regular CPDN boards for awhile now and I was expecting about a year for the deadlines here. The models take a long time and some of the experiments (most notably the main/current one) don't really need the results quickly, but do need lots of them. Some of the other experiments may need shorter deadlines, for example the THC slowdown experiment is wrapping up now even though it only started about 6 months ago (if I remeber correctly). If it had been running on BOINC I would have expected it to have 2 to 3 month deadlines.
<br>John Keck
BOINCing since 2002/12/08
ID: 177 · Report as offensive     Reply Quote
old_user1
Avatar

Send message
Joined: 5 Aug 04
Posts: 907
Credit: 299,864
RAC: 0
Message 179 - Posted: 6 Aug 2004, 12:33:52 UTC - in response to Message 177.  

yes, the hadsm3 is a long-running experiment, experiments such as THC and the next one (sulphur cycle) will be shorter report deadlines. we've had a few people finish after 6-9 months, just imagine a not-so-fast computer that they only use 8 hours a day and you can easily see how it gets to 9 months at least!



ID: 179 · Report as offensive     Reply Quote
old_user355

Send message
Joined: 7 Aug 04
Posts: 187
Credit: 44,163
RAC: 0
Message 277 - Posted: 7 Aug 2004, 6:06:17 UTC - in response to Message 69.  

&gt; I did not think you could use Boinc 4.xx for seti yet??? On the other hand I
&gt; know your problem will happen when we can use them all.

You can't.

It gives you this message.

Message from server: To participate in this project, you must use major version 3 of the BOINC core client. Your core client is major version 4.

Same from Predictor.
ID: 277 · Report as offensive     Reply Quote

Message boards : Number crunching : BOINC user preferences

©2024 cpdn.org