Advanced search

Forums : Wish list : Better fpops estimates?
Message board moderation

To post messages, you must log in.

1 · 2 · Next

AuthorMessage
Profile KSMarksPsych
Volunteer tester
Avatar

Send message
Joined: 22 May 07
Posts: 92
Credit: 57,682
RAC: 0
Message 39 - Posted: 7 Jun 2007, 22:00:39 UTC

When you generate the next batch of work, can you reduce the fpops estimates?

I've only run 3 or 4 WUs, so my DCF is still quite close to one. The WUs are taking less than a minute on my computer (a P4 @2.8 that's probably doing some thermal throttling) but they come with an initial guess of about 15 minutes. After people run a bunch, DCF is going to drop. In the future, if WUs really do get longer, people will download bunches and bunches and then run the risk of late work.

A decrease by a 10 (or a bit less) should help.
Kathryn :o)
The BOINC FAQ Service
The Unofficial BOINC Wiki
The Trac System
More BOINC information than you can shake a stick of RAM at.
ID: 39 · Report as offensive     Reply Quote
Profile KSMarksPsych
Volunteer tester
Avatar

Send message
Joined: 22 May 07
Posts: 92
Credit: 57,682
RAC: 0
Message 46 - Posted: 8 Jun 2007, 2:40:51 UTC
Last modified: 8 Jun 2007, 2:47:11 UTC

My DCF seems to have stabilized at about .04 after about 200 results.

[edit]
It's this host.
[/edit]
Kathryn :o)
The BOINC FAQ Service
The Unofficial BOINC Wiki
The Trac System
More BOINC information than you can shake a stick of RAM at.
ID: 46 · Report as offensive     Reply Quote
John McLeod VII
Volunteer tester
Avatar

Send message
Joined: 8 Jun 07
Posts: 28
Credit: 305,469
RAC: 0
Message 47 - Posted: 8 Jun 2007, 2:43:52 UTC - in response to Message 46.  

My DCF seems to have stabilized at about .04 after about 200 results.

If all machines that have returned about 200 results are in the .04 range, a factor of 25 division of the fpops_est is in order.


BOINC WIKI
ID: 47 · Report as offensive     Reply Quote
Profile KSMarksPsych
Volunteer tester
Avatar

Send message
Joined: 22 May 07
Posts: 92
Credit: 57,682
RAC: 0
Message 48 - Posted: 8 Jun 2007, 2:47:53 UTC - in response to Message 47.  

My DCF seems to have stabilized at about .04 after about 200 results.

If all machines that have returned about 200 results are in the .04 range, a factor of 25 division of the fpops_est is in order.


Ouch... my mental math above was off.

See... this is why we love you John.

:)
Kathryn :o)
The BOINC FAQ Service
The Unofficial BOINC Wiki
The Trac System
More BOINC information than you can shake a stick of RAM at.
ID: 48 · Report as offensive     Reply Quote
Profile [AF>Futura Sciences>Linux] Thrr-Gilag
Volunteer tester
Avatar

Send message
Joined: 22 May 07
Posts: 32
Credit: 145,576
RAC: 0
Message 62 - Posted: 8 Jun 2007, 11:44:05 UTC
Last modified: 8 Jun 2007, 11:45:00 UTC

Even my PII has a very small DCF(0.058887)
My P4's DCF is, just like others, in the range of 0.04
------
Thrr-Gilag Kee'rr

L'Alliance Francophone
ID: 62 · Report as offensive     Reply Quote
John McLeod VII
Volunteer tester
Avatar

Send message
Joined: 8 Jun 07
Posts: 28
Credit: 305,469
RAC: 0
Message 64 - Posted: 8 Jun 2007, 11:48:30 UTC - in response to Message 62.  
Last modified: 8 Jun 2007, 11:57:07 UTC

Even my PII has a very small DCF(0.058887)
My P4's DCF is, just like others, in the range of 0.04

OK, this makes the range of likely divisors 16 to 25.

[edit]

I have one host that has a DCF around 0.034 making the range 16 to 30. At this point, I would suggest dividing the fpops est by 25. I believe that this would give a range of DCF values of 0.5 to 2.0 which is just about perfect.


BOINC WIKI
ID: 64 · Report as offensive     Reply Quote
Profile Scott
Volunteer moderator
Project administrator
Project developer
Avatar

Send message
Joined: 1 Apr 07
Posts: 662
Credit: 13,742
RAC: 0
Message 74 - Posted: 8 Jun 2007, 14:24:11 UTC

I've divided the fpops estimate by a factor of 25. Tell me if you still think the estimate is way off.
Scott Kruger
Project Administrator, Cosmology@Home
ID: 74 · Report as offensive     Reply Quote
John McLeod VII
Volunteer tester
Avatar

Send message
Joined: 8 Jun 07
Posts: 28
Credit: 305,469
RAC: 0
Message 86 - Posted: 8 Jun 2007, 16:13:21 UTC - in response to Message 74.  

I've divided the fpops estimate by a factor of 25. Tell me if you still think the estimate is way off.

Of course that requires the ability to fetch work...


BOINC WIKI
ID: 86 · Report as offensive     Reply Quote
John McLeod VII
Volunteer tester
Avatar

Send message
Joined: 8 Jun 07
Posts: 28
Credit: 305,469
RAC: 0
Message 94 - Posted: 8 Jun 2007, 18:58:05 UTC - in response to Message 74.  

I've divided the fpops estimate by a factor of 25. Tell me if you still think the estimate is way off.

Of course, if you start sending multiple low level tasks in a single BOINC task, don't forget to multiply the fpops estimate by the number you package up together.


BOINC WIKI
ID: 94 · Report as offensive     Reply Quote
Profile Scott
Volunteer moderator
Project administrator
Project developer
Avatar

Send message
Joined: 1 Apr 07
Posts: 662
Credit: 13,742
RAC: 0
Message 95 - Posted: 8 Jun 2007, 19:17:28 UTC - in response to Message 94.  

I've divided the fpops estimate by a factor of 25. Tell me if you still think the estimate is way off.

Of course, if you start sending multiple low level tasks in a single BOINC task, don't forget to multiply the fpops estimate by the number you package up together.

Of course.
Scott Kruger
Project Administrator, Cosmology@Home
ID: 95 · Report as offensive     Reply Quote
John McLeod VII
Volunteer tester
Avatar

Send message
Joined: 8 Jun 07
Posts: 28
Credit: 305,469
RAC: 0
Message 98 - Posted: 8 Jun 2007, 23:36:57 UTC - in response to Message 95.  

I've divided the fpops estimate by a factor of 25. Tell me if you still think the estimate is way off.

Of course, if you start sending multiple low level tasks in a single BOINC task, don't forget to multiply the fpops estimate by the number you package up together.

Of course.

I am a Software Engineer myself, and I know how easy it is to forget something this simple :)


BOINC WIKI
ID: 98 · Report as offensive     Reply Quote
Profile Misfit
Volunteer tester
Avatar

Send message
Joined: 9 Jun 07
Posts: 150
Credit: 237,789
RAC: 0
Message 102 - Posted: 9 Jun 2007, 4:39:04 UTC - in response to Message 94.  

I've divided the fpops estimate by a factor of 25. Tell me if you still think the estimate is way off.

Of course, if you start sending multiple low level tasks in a single BOINC task, don't forget to multiply the fpops estimate by the number you package up together.

I'd love to see how BStudio reacts.
me@rescam.org
ID: 102 · Report as offensive     Reply Quote
Henri Tapani Heinonen
Volunteer tester
Avatar

Send message
Joined: 8 Jun 07
Posts: 2
Credit: 17,197
RAC: 0
Message 104 - Posted: 9 Jun 2007, 6:29:50 UTC

Hi everybody!

What is DCF? What is fpops_est? How to know what are the correct values for them? I am sorry for my basic questions. I just do not understand these things...

Henri Heinonen.
ID: 104 · Report as offensive     Reply Quote
Honza
Volunteer tester

Send message
Joined: 21 May 07
Posts: 26
Credit: 5,222,146
RAC: 0
Message 105 - Posted: 9 Jun 2007, 8:14:03 UTC

Henri,

I suggest visitng BOINC Wiki and spending some time reading the basics. Good reading.
(see the link in signatures).
ID: 105 · Report as offensive     Reply Quote
John McLeod VII
Volunteer tester
Avatar

Send message
Joined: 8 Jun 07
Posts: 28
Credit: 305,469
RAC: 0
Message 109 - Posted: 9 Jun 2007, 15:33:21 UTC - in response to Message 104.  

Hi everybody!

What is DCF? What is fpops_est? How to know what are the correct values for them? I am sorry for my basic questions. I just do not understand these things...

Henri Heinonen.

DCF is Duration Correction Factor. The ratio of what the client believes the work required is to what the server believes the work required is.

fpops_est is what the server believes the work required is.

fpops est is correct if the median DCF for projects that have returned enough work recently is 1.0. The fpops_est can be close enough if the median DCF is near 1. Problems start when the median DCF is 0.1 or below or 10.0 and above, and the problems get worse the further away from 1.0 that median is.


BOINC WIKI
ID: 109 · Report as offensive     Reply Quote
Profile KSMarksPsych
Volunteer tester
Avatar

Send message
Joined: 22 May 07
Posts: 92
Credit: 57,682
RAC: 0
Message 273 - Posted: 19 Jun 2007, 14:58:53 UTC

The ETA on the new work units for this host was 21 minutes. The actual run time is roughly 20 minutes.

So the fpops estimate scaled up well.

But my DCF is still extremely low at about .33.
Kathryn :o)
The BOINC FAQ Service
The Unofficial BOINC Wiki
The Trac System
More BOINC information than you can shake a stick of RAM at.
ID: 273 · Report as offensive     Reply Quote
John McLeod VII
Volunteer tester
Avatar

Send message
Joined: 8 Jun 07
Posts: 28
Credit: 305,469
RAC: 0
Message 277 - Posted: 19 Jun 2007, 20:47:06 UTC - in response to Message 273.  

The ETA on the new work units for this host was 21 minutes. The actual run time is roughly 20 minutes.

So the fpops estimate scaled up well.

But my DCF is still extremely low at about .33.

The Estimated Time that you see in the manager is based on the DCF.


BOINC WIKI
ID: 277 · Report as offensive     Reply Quote
Profile KSMarksPsych
Volunteer tester
Avatar

Send message
Joined: 22 May 07
Posts: 92
Credit: 57,682
RAC: 0
Message 278 - Posted: 19 Jun 2007, 22:44:50 UTC - in response to Message 277.  

The ETA on the new work units for this host was 21 minutes. The actual run time is roughly 20 minutes.

So the fpops estimate scaled up well.

But my DCF is still extremely low at about .33.

The Estimated Time that you see in the manager is based on the DCF.


I know.

I should have made my point better... sorry.

When the WUs were made longer the estimates were scaled up well. But they are still out of whack given the extremely low DCF.


You would think I would have learned by now not to post while tired. But nah. What's the fun in doing it while coherent.

:-)

Kathryn :o)
The BOINC FAQ Service
The Unofficial BOINC Wiki
The Trac System
More BOINC information than you can shake a stick of RAM at.
ID: 278 · Report as offensive     Reply Quote
Profile [AF>Futura Sciences>Linux] Thrr-Gilag
Volunteer tester
Avatar

Send message
Joined: 22 May 07
Posts: 32
Credit: 145,576
RAC: 0
Message 280 - Posted: 19 Jun 2007, 22:55:54 UTC

I have a strange thing for my DCF :

It is around 0.8 for my P4 and my Athlon computer, but dive around 0.2 for my Pentium 2 ^^

A problem with the power test I suppose (perhaps running when my ftp server was working).

So I thinck the fpops estimates is quite good for my concern.
------
Thrr-Gilag Kee'rr

L'Alliance Francophone
ID: 280 · Report as offensive     Reply Quote
Profile Scott
Volunteer moderator
Project administrator
Project developer
Avatar

Send message
Joined: 1 Apr 07
Posts: 662
Credit: 13,742
RAC: 0
Message 285 - Posted: 20 Jun 2007, 14:09:46 UTC

I've been tweaking it recently. It's now 3 times less than it was before (about 2e12), so we'll see how that works out.
Scott Kruger
Project Administrator, Cosmology@Home
ID: 285 · Report as offensive     Reply Quote
1 · 2 · Next

Forums : Wish list : Better fpops estimates?