Massive drop of credits per CPU hour

Message boards : Number crunching : Massive drop of credits per CPU hour
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4

AuthorMessage
frankhagen

Send message
Joined: 19 Aug 11
Posts: 76
Credit: 2,002,860
RAC: 0
Message 376 - Posted: 21 Nov 2011, 18:59:19 UTC - in response to Message 374.  

... fourth is credit based on runtime (a new one implemented recently by DA for the cases where creditnew doesn't work well)."

go read the docs...

Link to doc (or other source) for #4, please? I don't recognise it.



well, i don't read those docs - you'll have to ask Teemu what he meant..
ID: 376 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
zombie67 [MM]
Avatar

Send message
Joined: 19 Aug 11
Posts: 31
Credit: 63,078,601
RAC: 27,106
Message 377 - Posted: 22 Nov 2011, 3:55:24 UTC - in response to Message 369.  

55 credits for 36 hours. Oh how we laughed.


looks like you managed to draw a blank in credit-lotto.. :(


Quoth DA:

The goals of CreditNew involve long-term averages. It makes no promises about individual jobs or about credit/hour.


CreditNew = apparently random credits to individual users. IOW, be lucky or be screwed.
Reno, NV
Team: SETI.USA
ID: 377 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
ChertseyAl

Send message
Joined: 19 Aug 11
Posts: 45
Credit: 1,014,069
RAC: 0
Message 379 - Posted: 22 Nov 2011, 8:52:31 UTC

Hot news! Hold the front page! A new winner!

http://stat.la.asu.edu/NumberFields/result.php?resultid=573572

41 credits for 49 hours!

Rejoice! Rejoice! For CreditNew hath shown us The New Way! Let the first-born child of every family bring a goat for sacrifice to to sacred altar of Fairness!

Al.
ID: 379 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
ChertseyAl

Send message
Joined: 19 Aug 11
Posts: 45
Credit: 1,014,069
RAC: 0
Message 380 - Posted: 22 Nov 2011, 8:54:20 UTC - in response to Message 377.  


CreditNew = apparently random credits to individual users. IOW, be lucky or be screwed.


I know which group I fall into :)

Al.
ID: 380 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Senilix

Send message
Joined: 24 Aug 11
Posts: 1
Credit: 123,254
RAC: 0
Message 381 - Posted: 22 Nov 2011, 12:46:04 UTC - in response to Message 380.  

@Al

All of your WUs with improper (too high/too low) credit have one thing in common: all show up with a run time of 0.00 sec.

I would guess this is (part of) the problem!
ID: 381 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
ChertseyAl

Send message
Joined: 19 Aug 11
Posts: 45
Credit: 1,014,069
RAC: 0
Message 382 - Posted: 22 Nov 2011, 13:24:45 UTC - in response to Message 381.  


All of your WUs with improper (too high/too low) credit have one thing in common: all show up with a run time of 0.00 sec.


Ah, I hadn't made that connection. It's odd that some WUs show zero, others don't. I've used a dozen different hosts, but ATM I'm using 7 similar ones all running BOINC 5.10.45.

I'll see if I can find out what's going on, though I suspect it is the WUs themsleves rather than the hosts. Thanks for the heads-up :)

Al.
ID: 382 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
frankhagen

Send message
Joined: 19 Aug 11
Posts: 76
Credit: 2,002,860
RAC: 0
Message 383 - Posted: 22 Nov 2011, 13:28:09 UTC - in response to Message 381.  

@Al

All of your WUs with improper (too high/too low) credit have one thing in common: all show up with a run time of 0.00 sec.

I would guess this is (part of) the problem!


good point!

older boinc-clients to not ptoperly report run time and cpu time. looks like creditnew has fallen into another pit dug out in boinc-central..
ID: 383 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
ChertseyAl

Send message
Joined: 19 Aug 11
Posts: 45
Credit: 1,014,069
RAC: 0
Message 384 - Posted: 22 Nov 2011, 13:35:13 UTC - in response to Message 383.  


older boinc-clients to not ptoperly report run time and cpu time. looks like creditnew has fallen into another pit dug out in boinc-central..


Yes, I got bitten on the a4se by another project like that. Hmmm. I'll screen-scrape all of my results into Excel and look for the pattern.

One thing's for sure though, I'm not installing BOINC 6.x on anything else. I have to use it on a couple of machines and I hate it :(

Al.
ID: 384 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Eric Driver
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 8 Jul 11
Posts: 1318
Credit: 403,864,778
RAC: 289,239
Message 385 - Posted: 22 Nov 2011, 17:35:31 UTC - in response to Message 381.  

@Al

All of your WUs with improper (too high/too low) credit have one thing in common: all show up with a run time of 0.00 sec.

I would guess this is (part of) the problem!


Good catch Senilix.

The app also measures its own elapsed time internally, and this agrees with Al's cpu time.
ID: 385 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
ChertseyAl

Send message
Joined: 19 Aug 11
Posts: 45
Credit: 1,014,069
RAC: 0
Message 386 - Posted: 22 Nov 2011, 18:05:52 UTC - in response to Message 385.  

It's not *quite* as simple as it seems though. I've grabbed all of my results and put them in a spreadsheet so that I can sort/filter them.

Observations:

1) Those with spectacularly poor credit have a zero run time, those are four V2.02 and one V2.03 (5 WUs)

2) Of those with a zero run time (21 WUs), 5 have low credit (see above), 1 has 'sensible' credit, 10 have BIG credit and 5 have *OMFG* credit. All of these are V2.02 but one. Seems that zero run time is responsible for random credit rather than CreditNew (can't believe I'm saying that!).

3) CreditNew seems to have actually settled down. Credits were higher when I started around 17th/18th and seem to level out a day of so later. Maybe CreditNew works? (*stabs self in eye with fork*). Later WUs with non-zero run times seem to give pretty constitent credit (*pinches self*).

I can, as yet see no other patterns. Would love to able to predict if my current long-runners are going to be winners or losers in the lottery :)

Al.
ID: 386 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
ChertseyAl

Send message
Joined: 19 Aug 11
Posts: 45
Credit: 1,014,069
RAC: 0
Message 387 - Posted: 22 Nov 2011, 18:08:49 UTC - in response to Message 386.  

Oh yeah, sample size was 232 WUs. Removed 2 from my set of 234 as they were <1 second and didn't really fit in with anything (rounding errors etc).

Also, thumbs up to @Senilix for spotting the underlying problem :)

Al.
ID: 387 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
frankhagen

Send message
Joined: 19 Aug 11
Posts: 76
Credit: 2,002,860
RAC: 0
Message 389 - Posted: 22 Nov 2011, 18:53:06 UTC - in response to Message 387.  

Also, thumbs up to @Senilix for spotting the underlying problem :)


YUP!

what's rock proof by now:

creditnew is using elapsed time not cpu time - that's where you earned your merits.
this was introduced probably when AQUA showed up with MT-apps and the client reported cpu time of a single core. crazy enough, because elapsed time can be about everything + cpu time. just think about virtual machines..


creditnew is still using those faulty benchmarks to calculate credits, so it's totally open for cheating.


creditnew will not work for projects that have WU's with largely various runtimes. it will simply continue to grant random credits and send way too much or way too few WU's and by that foul up everything hostside.



ID: 389 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Rasputin42

Send message
Joined: 5 Nov 11
Posts: 25
Credit: 1,450,603
RAC: 0
Message 396 - Posted: 24 Nov 2011, 0:16:12 UTC

Hi,
As this project does not grand above normal credits, i am out of here.
Good look with participants, that are simply interested in the project.
Given equal credits, i rather crunch other projects.
This one is not that interesting.
ID: 396 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Eric Driver
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 8 Jul 11
Posts: 1318
Credit: 403,864,778
RAC: 289,239
Message 398 - Posted: 24 Nov 2011, 18:14:22 UTC - in response to Message 373.  

Not sure how to "get rid of it", as it's embedded in the latest server code.


yes it is - the default.

but, taken from the MOOwarp boards:

"There's actually four credit systems* in the BOINC validator code by default. First is that dynamic scheme for "device neutrality", second one is to get credit from WU (which is what we use), third is to grant no credit (instead trickle calls can grant credit) during validation and fourth is credit based on runtime (a new one implemented recently by DA for the cases where creditnew doesn't work well)."

go read the docs...



I went ahead and implemented the "--credit_from_runtime X" option in the validator. Now we wait to see if this fixes the problem with CreditNew...
ID: 398 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Richard Haselgrove

Send message
Joined: 28 Oct 11
Posts: 179
Credit: 220,442,182
RAC: 128,749
Message 410 - Posted: 2 Dec 2011, 17:58:26 UTC

Funny how quiet it gets in here when credit is going back up again :-)


(Direct link)
ID: 410 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
ChertseyAl

Send message
Joined: 19 Aug 11
Posts: 45
Credit: 1,014,069
RAC: 0
Message 581 - Posted: 23 Mar 2012, 19:16:26 UTC - in response to Message 398.  


I went ahead and implemented the "--credit_from_runtime X" option in the validator. Now we wait to see if this fixes the problem with CreditNew...


Does this mean that my 5.10.45 clients will now get sensible credit? If so, I'll start crunching this project again :)

Al.
ID: 581 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
frankhagen

Send message
Joined: 19 Aug 11
Posts: 76
Credit: 2,002,860
RAC: 0
Message 582 - Posted: 23 Mar 2012, 20:10:28 UTC - in response to Message 581.  


I went ahead and implemented the "--credit_from_runtime X" option in the validator. Now we wait to see if this fixes the problem with CreditNew...


Does this mean that my 5.10.45 clients will now get sensible credit? If so, I'll start crunching this project again :)


at least i think so - give it a try.

oh - and be careful not to take the real long WU's for the slow hosts!

ID: 582 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
ChertseyAl

Send message
Joined: 19 Aug 11
Posts: 45
Credit: 1,014,069
RAC: 0
Message 583 - Posted: 23 Mar 2012, 20:17:48 UTC - in response to Message 582.  

be careful not to take the real long WU's for the slow hosts!


Yeah, I'll go for bounded ones initially :)

Al.
ID: 583 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Eric Driver
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 8 Jul 11
Posts: 1318
Credit: 403,864,778
RAC: 289,239
Message 585 - Posted: 24 Mar 2012, 6:07:03 UTC - in response to Message 581.  


I went ahead and implemented the "--credit_from_runtime X" option in the validator. Now we wait to see if this fixes the problem with CreditNew...


Does this mean that my 5.10.45 clients will now get sensible credit? If so, I'll start crunching this project again :)

Al.


Sorry, I've been awol the whole day.

Yes, the credit problem was fixed several months ago. I have had no complaints since then.
ID: 585 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
ChertseyAl

Send message
Joined: 19 Aug 11
Posts: 45
Credit: 1,014,069
RAC: 0
Message 599 - Posted: 9 Apr 2012, 17:45:09 UTC - in response to Message 585.  

Yes, the credit problem was fixed several months ago. I have had no complaints since then.


No complaints here either - Predictable credit on every WU so far :)

It's a pity that the estimated runtimes of the bounded WUs are way too high, and actual runtimes are so variable, makes it difficult to keep full caches. At least it's worth crunching them now though :)

Cheers,

Al.
ID: 599 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Previous · 1 · 2 · 3 · 4

Message boards : Number crunching : Massive drop of credits per CPU hour


Main page · Your account · Message boards


Copyright © 2024 Arizona State University