Advanced search

Message boards : Announcements : Longer WUs

Author Message
Profile Scott
Volunteer moderator
Project administrator
Project developer
Avatar
Send message
Joined: 1 Apr 07
Posts: 662
Credit: 13,742
RAC: 0
Message 1142 - Posted: 19 Jul 2007, 15:11:29 UTC
Last modified: 19 Jul 2007, 21:04:37 UTC

Longer WUs are going to be here very shortly, along with new versions of CAMB that support checkpointing. There are a couple of things to note about the new system:

- CAMB will produce a couple of checkpoint files that are somewhat large (greater than 200 Mb in the final stages). These files are deleted at the end of the computation, so there's no need to worry.
- Longer WUs with the checkpointing system take over 3 hours on reference computer, so they may take over 12 hours to finish on some slower systems.
- The new CAMB version (1.22) will be released for all platforms currently supported. The checkpointing system is very new and will almost certainly have some problems, so expect invalidated results to jump for a while until we figure out the problems.

EDIT:

Some more notes:

-No more of the short WUs are being made, but at the time of this post (7/19 @ 4:03 PM CST), there are still over 20k pending results, waiting to be crunched. Once most of these are finished, longer WUs will be sent out.
-The maximum number of WUs in progress at one time has been lowered to 25 in anticipation of the longer WUs. This will temporarily disrupt the flow of short WUs, but everything should work fine as soon as they are finished.
____________
Scott Kruger
Project Administrator, Cosmology@Home

Profile Jayargh
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 25 Jun 07
Posts: 508
Credit: 2,282,158
RAC: 0
Message 1144 - Posted: 19 Jul 2007, 15:22:44 UTC

Scott will a working progress bar be included in camb 1.22?

Profile Scott
Volunteer moderator
Project administrator
Project developer
Avatar
Send message
Joined: 1 Apr 07
Posts: 662
Credit: 13,742
RAC: 0
Message 1145 - Posted: 19 Jul 2007, 15:39:29 UTC - in response to Message 1144.

Scott will a working progress bar be included in camb 1.22?

Not yet.
____________
Scott Kruger
Project Administrator, Cosmology@Home

Svenie25
Send message
Joined: 21 Jun 07
Posts: 68
Credit: 34,092
RAC: 0
Message 1146 - Posted: 19 Jul 2007, 16:52:59 UTC - in response to Message 1145.

Scott will a working progress bar be included in camb 1.22?

Not yet.


I think, this will make some people nervous. ;) But lets see, whats going up with the new WU´s. Don´t forget to adjust the credits.
____________
MfG
Sven S. aus G.

Nvgnte
Avatar
Send message
Joined: 24 Jun 07
Posts: 49
Credit: 186,805
RAC: 7
Message 1147 - Posted: 19 Jul 2007, 17:18:28 UTC
Last modified: 19 Jul 2007, 17:20:04 UTC

Yup, I have a P4-1800 Mhz that surely will crunch them on 12 hours (or even more)... at this stage, progress indicator is somehow necesary to check if WU is running correctly (and not stucked on an endless loop, for instance)

Anyway, time to time... this is an alpha phase so, it'll come one day or another ;)
____________
La Tierra de un Dios que no supo aceptar / su falso derecho a la libertad - Tierra Santa
Descárgate mi primer eBook Amaneceres

Profile teemac
Avatar
Send message
Joined: 3 Jul 07
Posts: 13
Credit: 749,995
RAC: 0
Message 1175 - Posted: 20 Jul 2007, 0:58:58 UTC

Just a quick comment for those with slower machines - It might be a good idea to go into your General Preferences and change your work cache to .1 days.

That way you won't be flooded with what BOINC thinks are the old short work units when they are really the new several hour long units. This would only be neccessary if you are crunching multiple projects. That way you won't miss any deadlines on other projects because of the extra time taken.

BOINC will sort it out over the first few hours of new WU's.
____________

Profile Misfit
Volunteer tester
Avatar
Send message
Joined: 9 Jun 07
Posts: 150
Credit: 237,789
RAC: 0
Message 1338 - Posted: 21 Jul 2007, 19:03:56 UTC - in response to Message 1147.

Yup, I have a P4-1800 Mhz that surely will crunch them on 12 hours (or even more)

My 3.4 Prescott estimates 6 hours.
____________
me@rescam.org

Profile [B@H] Ray
Volunteer tester
Send message
Joined: 22 May 07
Posts: 24
Credit: 7,688
RAC: 0
Message 1339 - Posted: 21 Jul 2007, 19:29:39 UTC - in response to Message 1175.

Just a quick comment for those with slower machines - It might be a good idea to go into your General Preferences and change your work cache to .1 days.

That way you won't be flooded with what BOINC thinks are the old short work units when they are really the new several hour long units. This would only be neccessary if you are crunching multiple projects. That way you won't miss any deadlines on other projects because of the extra time taken.

BOINC will sort it out over the first few hours of new WU's.

Looks like the time was adjusted on them at the server so we don't get a lot of units.
____________


Pizza@Home Rays Place Rays place Forums

Profile [AF>Futura Sciences>Linux] Thrr-Gilag
Volunteer tester
Avatar
Send message
Joined: 22 May 07
Posts: 32
Credit: 145,576
RAC: 0
Message 1378 - Posted: 22 Jul 2007, 18:51:44 UTC

Just one question on checkpoint : How often are save the state of calculation ?
____________
------
Thrr-Gilag Kee'rr

L'Alliance Francophone

Profile Jayargh
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 25 Jun 07
Posts: 508
Credit: 2,282,158
RAC: 0
Message 1381 - Posted: 22 Jul 2007, 19:02:04 UTC - in response to Message 1378.

Just one question on checkpoint : How often are save the state of calculation ?



I beleive 120seconds-2 minutes
____________

Profile Ray Murray
Avatar
Send message
Joined: 28 Jun 07
Posts: 12
Credit: 44,336
RAC: 0
Message 1382 - Posted: 22 Jul 2007, 19:21:45 UTC

From this post from Scott, the checkpoints are internal and aren't through Boinc so the wus will run to completion (or failure) before Boinc will switch to another project.

Profile [AF>Futura Sciences>Linux] Thrr-Gilag
Volunteer tester
Avatar
Send message
Joined: 22 May 07
Posts: 32
Credit: 145,576
RAC: 0
Message 1399 - Posted: 23 Jul 2007, 6:57:52 UTC - in response to Message 1382.

From this post from Scott, the checkpoints are internal and aren't through Boinc so the wus will run to completion (or failure) before Boinc will switch to another project.

No, if the option "leave application in memory" on boinc 5.8.16, the Wu is not done in one shot.
____________
------
Thrr-Gilag Kee'rr

L'Alliance Francophone

Profile vajras
Send message
Joined: 30 Jun 07
Posts: 9
Credit: 100,954
RAC: 0
Message 1401 - Posted: 23 Jul 2007, 8:32:56 UTC

So ... may i ask when exactly are you planning to implement the "progress bar"?

The way i read this is.. longer unspecified time work units, together with a not-so-visible-to-us checkpoint system at an unspecified arbitary final stage will most likely make a few people more than a little nervous.

And is it also possible to give an indication of exactly what sort of an invalidated result prob we may expect?

Many thanks

Profile Wabbit98
Send message
Joined: 31 Jul 07
Posts: 11
Credit: 35,800
RAC: 0
Message 1929 - Posted: 10 Aug 2007, 17:40:12 UTC

As long as you have a checkpoint in place I do not mind if the progress bar is not ready yet. I remember crunching for RCN before checkpointing and really long WU's. Also BURP does not have checkpointing and a variety of length WU's, but really short deadlines.

Ni

Profile MALCOLM D BARKER
Send message
Joined: 26 Aug 07
Posts: 3
Credit: 9,390
RAC: 0
Message 2525 - Posted: 6 Sep 2007, 3:15:32 UTC - in response to Message 1142.

Uh, what happens if I don't complete a task by the report deadline, does that make me a bad cruncher? How can I set cosmolegy@home so it runs with something else so I get a screensaver?
Thanks

Longer WUs are going to be here very shortly, along with new versions of CAMB that support checkpointing. There are a couple of things to note about the new system:

- CAMB will produce a couple of checkpoint files that are somewhat large (greater than 200 Mb in the final stages). These files are deleted at the end of the computation, so there's no need to worry.
- Longer WUs with the checkpointing system take over 3 hours on reference computer, so they may take over 12 hours to finish on some slower systems.
- The new CAMB version (1.22) will be released for all platforms currently supported. The checkpointing system is very new and will almost certainly have some problems, so expect invalidated results to jump for a while until we figure out the problems.

EDIT:

Some more notes:

-No more of the short WUs are being made, but at the time of this post (7/19 @ 4:03 PM CST), there are still over 20k pending results, waiting to be crunched. Once most of these are finished, longer WUs will be sent out.
-The maximum number of WUs in progress at one time has been lowered to 25 in anticipation of the longer WUs. This will temporarily disrupt the flow of short WUs, but everything should work fine as soon as they are finished.


____________

Profile Jayargh
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 25 Jun 07
Posts: 508
Credit: 2,282,158
RAC: 0
Message 2530 - Posted: 6 Sep 2007, 4:44:17 UTC - in response to Message 2525.

Malcom in the future if you have a question please start a new thread in general or technical not in announcements if you cant find an answer anywhere else.

Uh, what happens if I don't complete a task by the report deadline, does that make me a bad cruncher?


No the workunit will be sent to someone else and the total allowed to you will be reduced by 1 for everyone not returned.

How can I set cosmolegy@home so it runs with something else so I get a screensaver?
Thanks


Add another project in your Boinc manager then set that projects resource share to run 50/50 with C@H so 1 core runs cosmo and 1 core runs another project with screensaver.
____________

Message boards : Announcements : Longer WUs