I don't know for sure either. As I said, I never tested it. But it sounds at least possible to me.
@Ageless: I don't think fumbling with the clock would make any difference in reported crunch times. Daylight savings never did for me (only effect that ever had was a coupe of "no heartbeat from core client" errors) I think the CPU seconds are calculated completely independently from wall clock time...
Has anyone actually thought about PMing this person to get it from the horse's mouth, so to speak? I wouldn't even know what questions to ask, but someone on this thread obviously does. If this person has a logical explanation, then the mystery can be laid to rest. If not, then maybe that's the time to look into it further.
EDIT: From his account page, it looks like he at least currently lives in Germany. So presumably a German-speaker would need to contact him.
XJR-Maniac just PMed me back and explained what happened. I'm translating this from German as I write so I hope it will be understandable...
Apparently, the theory about having some more boxes without internet access was pretty much right. XJR-Maniac wrote he is running Einstein on a couple of company PCs but it is restricted from connecting by a firewall. So he puts the results on a USB flash drive, crunches them and takes them back home (pretty much as we figured). By loading the WUs for many clients on one PC he causes duplicate clients (dunno how, he didn't elaborate very much). He merges those as quickly as possible because he thinks it's somehow "neater" (he has made the experience that sometimes he couldn't merge duplicate clients if he waits too long), but this triggers a "resend lost result" event from the scheduler when the remaining host connects the next time. When he uploads the results from the other clients, the computation time is shown as the time between resend and upload.
I don't know for sure either.
)
I don't know for sure either. As I said, I never tested it. But it sounds at least possible to me.
@Ageless: I don't think fumbling with the clock would make any difference in reported crunch times. Daylight savings never did for me (only effect that ever had was a coupe of "no heartbeat from core client" errors) I think the CPU seconds are calculated completely independently from wall clock time...
Has anyone actually thought
)
Has anyone actually thought about PMing this person to get it from the horse's mouth, so to speak? I wouldn't even know what questions to ask, but someone on this thread obviously does. If this person has a logical explanation, then the mystery can be laid to rest. If not, then maybe that's the time to look into it further.
EDIT: From his account page, it looks like he at least currently lives in Germany. So presumably a German-speaker would need to contact him.
Since I am a German native
)
Since I am a German native speaker, I just wrote the guy a PM. I'll keep you posted.
XJR-Maniac just PMed me back
)
XJR-Maniac just PMed me back and explained what happened. I'm translating this from German as I write so I hope it will be understandable...
Apparently, the theory about having some more boxes without internet access was pretty much right. XJR-Maniac wrote he is running Einstein on a couple of company PCs but it is restricted from connecting by a firewall. So he puts the results on a USB flash drive, crunches them and takes them back home (pretty much as we figured). By loading the WUs for many clients on one PC he causes duplicate clients (dunno how, he didn't elaborate very much). He merges those as quickly as possible because he thinks it's somehow "neater" (he has made the experience that sometimes he couldn't merge duplicate clients if he waits too long), but this triggers a "resend lost result" event from the scheduler when the remaining host connects the next time. When he uploads the results from the other clients, the computation time is shown as the time between resend and upload.