Message boards : Number crunching : @Markku - Estimated WU Completion
Message board moderation

To post messages, you must log in.

Previous · 1 · 2

AuthorMessage
John McLeod VII
Avatar

Send message
Joined: 2 Sep 04
Posts: 165
Credit: 146,925
RAC: 0
Message 8987 - Posted: 28 Jul 2005, 2:40:32 UTC - in response to Message 8978.  

> Hmm, not sure i understand.
> The variable number of turns are tuned in the input files, and the time
> estimate(well, actually the fops estimate) is based on the number of turns we
> ask to be done.
>
And the correction factor does its work based on the original (fpops) estimate from the developers, so the estimates that are proportional to some known input value (like LHC has) are no problem at all. The real problem is the projects where the developers have no real idea what the running time is going to be because the inputs are too complex (like Protein Predictor has).

After a month or two of the new versions being out in general hands, you should have a good idea how wrong the initial estimates are (take the average correction factor, and you won't be too far off from the % error in the original fpops estimate).


BOINC WIKI
ID: 8987 · Report as offensive     Reply Quote
John McLeod VII
Avatar

Send message
Joined: 2 Sep 04
Posts: 165
Credit: 146,925
RAC: 0
Message 8988 - Posted: 28 Jul 2005, 2:44:26 UTC - in response to Message 8979.  

> > I am waiting for the next recommended release where BOINC "learns"
> reality.
> > The only fly in the ointment is LHC@Home with the variable number of
> turns.
> >
>
> The picture is more involved than this, given that any unit potentially can
> terminate in a few seconds or minutes if the beam is unstable. LHC@Home is not
> the only project with this sort of variable execution. MDIV units run by a
> closed Alpha project that I know Paul is aware of can vary according the the
> requirements of the input files. The project's home page suggests that time
> estimates may be 'wildly inaccurate'. What's worse, the MDIV units neither
> update their progress meters, nor write checkpoint data. I can't see how BOINC
> 4.7+ can learn anything effective from either SixTrack or MDIV
>
Yes, the sudden death problem was taken into account. This is precisely why the estimate is reduced slowly. If a completion time is less than 10% of the midified original estimate, then the estimate will move down even more slowly just so that a few fast returning results will not upset the estimate too much.


BOINC WIKI
ID: 8988 · Report as offensive     Reply Quote
Profile Paul D. Buck

Send message
Joined: 2 Sep 04
Posts: 545
Credit: 148,912
RAC: 0
Message 8996 - Posted: 28 Jul 2005, 10:05:10 UTC

John,

Ok, thanks ... I get it now ... :)

Then again I have not been at my best this last week (coming out of it??) and so, easily led astray.
ID: 8996 · Report as offensive     Reply Quote
Trulayne
Avatar

Send message
Joined: 22 Oct 04
Posts: 39
Credit: 46,748
RAC: 0
Message 9009 - Posted: 28 Jul 2005, 21:11:04 UTC

Any estimates as to when 4.7+ will be released to the general public or is it too soon to tell?



ID: 9009 · Report as offensive     Reply Quote
John McLeod VII
Avatar

Send message
Joined: 2 Sep 04
Posts: 165
Credit: 146,925
RAC: 0
Message 9010 - Posted: 28 Jul 2005, 22:56:19 UTC - in response to Message 9009.  

> Any estimates as to when 4.7+ will be released to the general public or is it
> too soon to tell?
>
I am a mushroom.


BOINC WIKI
ID: 9010 · Report as offensive     Reply Quote
Profile Paul D. Buck

Send message
Joined: 2 Sep 04
Posts: 545
Credit: 148,912
RAC: 0
Message 9012 - Posted: 28 Jul 2005, 23:41:18 UTC

I thought the line went: "I am the eggman, I am the walrus... koo koo ka choo"?
ID: 9012 · Report as offensive     Reply Quote
Henry Nebrensky

Send message
Joined: 13 Jul 05
Posts: 165
Credit: 14,925,288
RAC: 34
Message 9013 - Posted: 29 Jul 2005, 0:24:17 UTC - in response to Message 8907.  

> [edit: I just looked at [John's] hosts - oh my you are pushing the
> bounds of old and slow machines on BOINC!]

Well, I'm jealous - my old and slow machines are on RedHat 6.2, and I can't
find a working BOINC client :(

(I've compiled several, but the benchmarking always concludes that I have
zero, count 'em, ZERO processors installed. Sigh.)

Henry

ID: 9013 · Report as offensive     Reply Quote
John McLeod VII
Avatar

Send message
Joined: 2 Sep 04
Posts: 165
Credit: 146,925
RAC: 0
Message 9034 - Posted: 30 Jul 2005, 1:23:43 UTC - in response to Message 9012.  

> I thought the line went: "I am the eggman, I am the walrus... koo koo ka
> choo"?
>
Wrong line.

I am a mushroom, I must be a mushroom. They keep me in the dark and they feed me BullS****.


BOINC WIKI
ID: 9034 · Report as offensive     Reply Quote
Previous · 1 · 2

Message boards : Number crunching : @Markku - Estimated WU Completion


©2024 CERN