Too Many Work Units
13 Mar 2005 13:18:03 UTC
Topic 188503
(moderation:
Does anyone else have a problem with Einstein sending too many work units becuase it underestimates the amount of CPU time required to complete the unit. In my case, the CPU estimate is 23.5 hours when it actually takes over 30 hours of CPU. As a result, I am sent more work units than can be completed before the deadline.
SETI on the other hand OVERestimates the required CPU time. Estimate ia about 16 hours, actual requied time is about 10 hours.
Too Many Work Units
)
Yes, I have the same problem, my estimates are around 5-6 hours when it actually takes me 9-11 hours running in HT mode ... It would be alright if the Site had a 14 day deadline but with the 7 day deadline you have to be on the low side with your Connect Preference or you'll end up with to many WU's that can't be completed in time ...
> Does anyone else have a
)
> Does anyone else have a problem with Einstein sending too many work units
> becuase it underestimates the amount of CPU time required to complete the
> unit. In my case, the CPU estimate is 23.5 hours when it actually takes over
> 30 hours of CPU. As a result, I am sent more work units than can be completed
> before the deadline.
>
> SETI on the other hand OVERestimates the required CPU time. Estimate ia about
> 16 hours, actual requied time is about 10 hours.
>
Hi,
Another user Ben, has posted in Claimed credit with dual CPUs? what may be a possible solution for you. He has compilied a modified windows 4.19 BOINC client. If you are the adventurous type and are willing to take a risk of course.
whats the advantage of having
)
whats the advantage of having many WUs in cache ? you could just set "connect every 0.1 days" and you wont have any problems... and if one WU takes about 30 hours it wouldnt be too hard to connect to the internet every 30 hours would it ? or set the cache at lets say 1,2 days or so and you should get 2 WUs what also should be ok...
> or set the cache at lets
)
> or set the cache at lets say 1,2 days or so and you should get 2 WUs what
> also should be ok...
No, that's a mistake I also did once. You have to use 1.2 days, otherwise the ,2 is just thrown away. The english use the '.' and ',' the other way then us germans ;)
But with the core information you are right.
And regarding the calculation of expected CPU-use:
Einstein is the only project that calculates the amount correct on my 2 puters.
Predictor and Seti overestimate a lot, and CPDN takes about 200h to reach the correct conclusion ;)
Grüße vom Sänger
> Does anyone else have a
)
> Does anyone else have a problem with Einstein sending too many work units
> becuase it underestimates the amount of CPU time required to complete the
> unit. In my case, the CPU estimate is 23.5 hours when it actually takes over
> 30 hours of CPU. As a result, I am sent more work units than can be completed
> before the deadline.
>
> SETI on the other hand OVERestimates the required CPU time. Estimate ia about
> 16 hours, actual requied time is about 10 hours.
>
I have one system that ran out of Seti WUs last weekend, so I attached Einstein to it. Einstein downloaded 8 WUs for it immediately. Later, when Seti came back up, I set the resource share as 2 to 1 - Seti over Einstein. Einstein has only downloaded 1 WU since the original 8 WUs (this is good), but the last two WUs from the original download have now timed out.
On the original download, Einstein should NOT have downloaded 8 WUs...it was too much, even for a 50-50 split between two projects.
I checked the two expired WUs and one has already had 3 results returned...I used Boincview to abort it. I used Boinc Manager to suspend Seti on that system, but later found out I could have used Boincview to do that too. Right now it's a race to finish the other WU before the fourth outstanding non-timed out result is returned. If I finish first and the other finishes before his time-out, we both get credit. If he finishes first and is validated before I get in...I get no credit.
Seti Classic Final Total: 11446 WU.