muon 1.07

Alles zum Projekt yoyo@home
Everything about the project yoyo@home
Nachricht
Autor
Benutzeravatar
yoyo
Vereinsvorstand
Vereinsvorstand
Beiträge: 8045
Registriert: 17.12.2002 14:09
Wohnort: Berlin
Kontaktdaten:

#37 Ungelesener Beitrag von yoyo » 21.06.2008 12:56

Bei weniger als 10 werden neue produziert. Alle 10min wird das geprüft.
yoyo
HILF mit im Rechenkraft-WiKi, dies gibts zu tun.
Wiki - FAQ - Verein - Chat

Bild Bild

Dennis Kautz
Vereinsmitglied
Vereinsmitglied
Beiträge: 4742
Registriert: 22.02.2003 02:12
Kontaktdaten:

#38 Ungelesener Beitrag von Dennis Kautz » 21.06.2008 12:56

yoyo hat geschrieben:Bei weniger als 10 werden neue produziert. Alle 10min wird das geprüft.
yoyo
Ok, danke!

Benutzeravatar
Beyond
Prozessor-Polier
Prozessor-Polier
Beiträge: 111
Registriert: 02.02.2008 01:48
Wohnort: Rum River watershed, MN, USA

Re: muon 1.07

#39 Ungelesener Beitrag von Beyond » 11.07.2008 16:58

I brought a new machine online a few days ago and have been keeping an eye on it. Today I noticed this:

http://www.rechenkraft.net/yoyo/workuni ... uid=878512

The result is marked as successful but appears to be denied credit because too many results were received?
Yet the machine returned the result only 4.5 hours after receiving it. Is there perhaps a problem in the credit awarding logic?

Thanks/Beyond

Benutzeravatar
yoyo
Vereinsvorstand
Vereinsvorstand
Beiträge: 8045
Registriert: 17.12.2002 14:09
Wohnort: Berlin
Kontaktdaten:

Re: muon 1.07

#40 Ungelesener Beitrag von yoyo » 11.07.2008 21:17

Hello,
the first result was sent on 6. July at 10:44. Deadline was 11. July 10:44. The result was not returned until the deadline. So Boinc created a second result and sent it to you on 11. July 11:03. But than at 15:16 the first result was returned and verified by Boinc. At 15:35 your result was returned which Boinc rejected, because there is already a valid result.
That is what happened. Honestly I do not know how to handle this in a different way. If I set the number of max results to 2 than also your result would be accepted, but your result would be compared with the first one and since muon is a simulation all results are different. Means the results would never be identical and the second result will be rejected as invalid.
yoyo
HILF mit im Rechenkraft-WiKi, dies gibts zu tun.
Wiki - FAQ - Verein - Chat

Bild Bild

Benutzeravatar
Beyond
Prozessor-Polier
Prozessor-Polier
Beiträge: 111
Registriert: 02.02.2008 01:48
Wohnort: Rum River watershed, MN, USA

Re: muon 1.07

#41 Ungelesener Beitrag von Beyond » 12.07.2008 06:39

I seem to remember other projects that have accepted multiple results with different outcomes. Can't quite remember which ones they were. Nanohive?
Another answer might be to not accept results that miss the due date.

Ananas
WU-Schieber
WU-Schieber
Beiträge: 1184
Registriert: 27.04.2008 18:37
Wohnort: Nordlichter Köln

Re: muon 1.07

#42 Ungelesener Beitrag von Ananas » 19.07.2008 16:40

Maybe an option would be not to accept the first result anymore.

Afterall that was the one that has caused the trouble, so if either of the two should not receive credits, it's sure the first.
vi BOINC/checkin_notes
:1,$s/bug/feature/g
:wq!

Erzaehlen sich Biologen eigentlich Klein-RNA-Witze?

Benutzeravatar
Beyond
Prozessor-Polier
Prozessor-Polier
Beiträge: 111
Registriert: 02.02.2008 01:48
Wohnort: Rum River watershed, MN, USA

Re: muon 1.07

#43 Ungelesener Beitrag von Beyond » 14.10.2008 00:33

> yoyo@home 10/13/2008 6:25:24 PM Message from server: No work sent
> yoyo@home 10/13/2008 6:25:24 PM Message from server: (reached per-host limit of 10 tasks)

My quads burn through 10 WUs pretty fast...

Benutzeravatar
yoyo
Vereinsvorstand
Vereinsvorstand
Beiträge: 8045
Registriert: 17.12.2002 14:09
Wohnort: Berlin
Kontaktdaten:

Re: muon 1.07

#44 Ungelesener Beitrag von yoyo » 14.10.2008 06:15

Hello,
I changed the limits now in the end phase of ogr to ensure a fast turn around of workunits. This can only be configured project wide, not per application. But I will change it back when ogr finished.
yoyo
HILF mit im Rechenkraft-WiKi, dies gibts zu tun.
Wiki - FAQ - Verein - Chat

Bild Bild

Benutzeravatar
Beyond
Prozessor-Polier
Prozessor-Polier
Beiträge: 111
Registriert: 02.02.2008 01:48
Wohnort: Rum River watershed, MN, USA

Re: muon 1.07

#45 Ungelesener Beitrag von Beyond » 14.10.2008 17:15

Thanks for the quick reply and good idea on the OGR front. Actually limits aren't necessarily a bad idea as long as they're not too tight...

Regards/Beyond

Seversen
Idle-Sammler
Idle-Sammler
Beiträge: 6
Registriert: 05.06.2008 20:23

Re: muon 1.07

#46 Ungelesener Beitrag von Seversen » 13.11.2008 16:46

Wondering is someone would know why this workunit ran for 100 hours, missed the deadline and still wasn't finished when I aborted it. http://www.rechenkraft.net/yoyo/result. ... id=1772335

Not sure if the unit was just too big for my P3 450mhz machine or if the unit had problems. I have run muons on this machine before with no problems.

Thanks.

Benutzeravatar
yoyo
Vereinsvorstand
Vereinsvorstand
Beiträge: 8045
Registriert: 17.12.2002 14:09
Wohnort: Berlin
Kontaktdaten:

Re: muon 1.07

#47 Ungelesener Beitrag von yoyo » 13.11.2008 19:20

Seversen hat geschrieben:Wondering is someone would know why this workunit ran for 100 hours, missed the deadline and still wasn't finished when I aborted it. http://www.rechenkraft.net/yoyo/result. ... id=1772335

Not sure if the unit was just too big for my P3 450mhz machine or if the unit had problems. I have run muons on this machine before with no problems.

Thanks.
To be honest, I never saw a muon wu running this long time. Since you aborted the wu, the results and logfiles are not uploaded to the Boinc server. Next time you should save the slot directory before aborting a wu for further analysis. But what I see in the stderr log of your wu it did run quite normal. This should be also shown in progress indicator, which was near by 90%.

Basically if a muon wu runs much longer than a muon wu before on a particular host it is on the way to generate a much better design than before.

yoyo
HILF mit im Rechenkraft-WiKi, dies gibts zu tun.
Wiki - FAQ - Verein - Chat

Bild Bild

Seversen
Idle-Sammler
Idle-Sammler
Beiträge: 6
Registriert: 05.06.2008 20:23

Re: muon 1.07

#48 Ungelesener Beitrag von Seversen » 13.11.2008 20:26

yoyo hat geschrieben:
Seversen hat geschrieben:Wondering is someone would know why this workunit ran for 100 hours, missed the deadline and still wasn't finished when I aborted it. http://www.rechenkraft.net/yoyo/result. ... id=1772335

Not sure if the unit was just too big for my P3 450mhz machine or if the unit had problems. I have run muons on this machine before with no problems.

Thanks.
To be honest, I never saw a muon wu running this long time. Since you aborted the wu, the results and logfiles are not uploaded to the Boinc server. Next time you should save the slot directory before aborting a wu for further analysis. But what I see in the stderr log of your wu it did run quite normal. This should be also shown in progress indicator, which was near by 90%.

Basically if a muon wu runs much longer than a muon wu before on a particular host it is on the way to generate a much better design than before.

yoyo

Thanks for the response. I'll keep that in mind for the next one that runs long.
But, it is hard to let a workunit continue to run when you know you won't get any credit for it since the deadline was up.

On the topic of deadlines, why is the deadline for the muon units 5 days? Definitely ended up being too short for my slow computer.

Thanks

Antworten

Zurück zu „Number crunching“