Message boards : Number crunching : Time to Completion
Message board moderation

To post messages, you must log in.

AuthorMessage
Pete49

Send message
Joined: 18 Sep 04
Posts: 35
Credit: 250,303
RAC: 0
Message 7957 - Posted: 4 Jun 2005, 6:28:11 UTC

Time to completion is being over estimated by almost a factor of 3 on all my crunchers.

Is this a deliberate effort to "spread" the work amongst the users or an error?


<img src="http://www.boincstats.com/stats/teambanner.php?teamname=GasBuddy"> <img src="http://www.boincstats.com/stats/banner.php?cpid=84c0cf7846cbf28338406e54b3eb8a83">
ID: 7957 · Report as offensive     Reply Quote
Profile sysfried

Send message
Joined: 27 Sep 04
Posts: 282
Credit: 1,415,417
RAC: 0
Message 7958 - Posted: 4 Jun 2005, 7:10:15 UTC - in response to Message 7957.  

> Time to completion is being over estimated by almost a factor of 3 on all my
> crunchers.
>
> Is this a deliberate effort to "spread" the work amongst the users or an
> error?
>
>
I can confirm this... time estimated... 10 hours... but WU's finish in aprox. 3 h

sysfried
ID: 7958 · Report as offensive     Reply Quote
Profile littleBouncer
Avatar

Send message
Joined: 23 Oct 04
Posts: 358
Credit: 1,439,205
RAC: 0
Message 7960 - Posted: 4 Jun 2005, 9:45:30 UTC

Same here; but I think it has to do with this quotes from the frontpage at 20.5.2005 12:39 UTC and 31.5.2005 10:12 UTC :

Some new scripts have been developed which hopefully should increase the speed and ease of use of the job submission. and
This should be resolved now and the <a>new scripts[/url] have been installed on the production server.


greetz littleBouncer

ID: 7960 · Report as offensive     Reply Quote
Profile Thierry Van Driessche
Avatar

Send message
Joined: 1 Sep 04
Posts: 157
Credit: 82,604
RAC: 0
Message 7973 - Posted: 5 Jun 2005, 14:19:46 UTC - in response to Message 7958.  

> I can confirm this... time estimated... 10 hours... but WU's finish in aprox.
> 3 h

Running Boinc v4.44 with HT CPU 2.4@2.88
estimated CPU time 13:05
crunched in some 4:50
ID: 7973 · Report as offensive     Reply Quote
Jayargh

Send message
Joined: 24 Oct 04
Posts: 79
Credit: 257,762
RAC: 0
Message 7974 - Posted: 5 Jun 2005, 14:27:54 UTC

It seems as I watch this phenomena on different runs as they have a formula for 100k runs and 1 for 1m runs that is calculated with your hosts benchmarks give you a consistent time to completion no matter what the parameters of a particular test ie: the time to completion calculated by boinc and lhc always come up about the same time on all my hosts but the REAL time changes dramatically mostly less but on occasion more and a few runs were almost spot on. Just an observation
ID: 7974 · Report as offensive     Reply Quote
Profile The Gas Giant

Send message
Joined: 2 Sep 04
Posts: 309
Credit: 715,258
RAC: 0
Message 7978 - Posted: 5 Jun 2005, 23:01:53 UTC

Why is it that the projects find it so hard to estimate the completion time like they do.

Currently the projects I attach to on my laptop have the following;

LHC Estimates 15hr15min Actual 4hr54min
PP@h Estimates 2hr32min Actual 1hr10min
S@h Estimates 6hr30min Actual 4hrs

My laptop doesn't get to attach to a network from 5pm Friday to 9am Monday and with the current wu schedular (V4.44) and estimated completion times I can't get sufficient work to last the weekend.

Hopefully LHC can do a little better with the next release of wu's. Come on folks, with the combined brainpower of the couple of physicists doing this project, surely this estimation can be better done (yes I know it's a function of the benchmarks, but it is something that can be played with on the project side as well).

Live long and crunch!

Paul
(S@H1 8888)
BOINC/SAH BETA
ID: 7978 · Report as offensive     Reply Quote
STE\/E

Send message
Joined: 2 Sep 04
Posts: 352
Credit: 1,393,150
RAC: 0
Message 7979 - Posted: 5 Jun 2005, 23:32:41 UTC

Why is it that the projects find it so hard to estimate the completion time like they do.
==========

Sometimes the Project itself over estimates the time it will take to complete the WU's in an effort to slow down the amount of WU's 1 Host can download at one time ...

But I don't know if this is the case with the latest LHC WU's or not, I know I can remember some time ago over at the Seti Site they purposefully added time to the WU's to slow down the downloads ...
ID: 7979 · Report as offensive     Reply Quote
Profile The Gas Giant

Send message
Joined: 2 Sep 04
Posts: 309
Credit: 715,258
RAC: 0
Message 7981 - Posted: 6 Jun 2005, 6:59:11 UTC - in response to Message 7979.  

> Sometimes the Project itself over estimates the time it will take to complete
> the WU's in an effort to slow down the amount of WU's 1 Host can download at
> one time ...
>
> But I don't know if this is the case with the latest LHC WU's or not, I know I
> can remember some time ago over at the Seti Site they purposefully added time
> to the WU's to slow down the downloads ...
>

Thanks PoorBoy, but the projects now have the maximum wu's per day setting, so overestimating the completion time is now redundant.....next!
ID: 7981 · Report as offensive     Reply Quote
Profile Markku Degerholm

Send message
Joined: 3 Sep 04
Posts: 212
Credit: 4,545
RAC: 0
Message 8003 - Posted: 7 Jun 2005, 21:06:37 UTC

I think the core clients computes the estimated computation directly from the "maximum number of integer/fp operations" value of workunit and benchmarked FLOPS/IOPS values of the host machine. The maximum operation count is very rough worst-case estimate. But if we find time to do some statistics to find correct values, maybe we will change them...

Markku Degerholm
LHC@home admin
ID: 8003 · Report as offensive     Reply Quote
Profile Ageless
Avatar

Send message
Joined: 18 Sep 04
Posts: 143
Credit: 27,645
RAC: 0
Message 8006 - Posted: 8 Jun 2005, 0:16:27 UTC

It's completely out of whack here. ;)

08/06/2005 00:15:49|LHC@home|Requesting 4.00 seconds of work
08/06/2005 00:15:51|LHC@home|Started download of wboinc7_v6s4hvnom__1__64.294_59.304__14_16__6__18_1_sixvf_boinc9645.zip
08/06/2005 00:15:52|LHC@home|Finished download of wboinc7_v6s4hvnom__1__64.294_59.304__14_16__6__18_1_sixvf_boinc9645.zip
08/06/2005 00:15:52|LHC@home|Throughput 60788 bytes/sec

I am getting another 4 hour unit. (It's claiming 14 hours at the start, but they run in 4 hours).
Jord

BOINC FAQ Service
ID: 8006 · Report as offensive     Reply Quote
Profile The Gas Giant

Send message
Joined: 2 Sep 04
Posts: 309
Credit: 715,258
RAC: 0
Message 8007 - Posted: 8 Jun 2005, 2:31:15 UTC - in response to Message 8003.  

> I think the core clients computes the estimated computation directly from the
> "maximum number of integer/fp operations" value of workunit and benchmarked
> FLOPS/IOPS values of the host machine. The maximum operation count is very
> rough worst-case estimate. But if we find time to do some statistics to find
> correct values, maybe we will change them...
>
Markku,

Even if you can have a quick look to try and reduce the estimated completion time to half of the current values it would go along way to overcoming the problem. Unless I'm mistaken, estimated wu completion times are more than double all platforms actual completion times.

Live long and crunch!

Paul.
ID: 8007 · Report as offensive     Reply Quote
Profile Paul D. Buck

Send message
Joined: 2 Sep 04
Posts: 545
Credit: 148,912
RAC: 0
Message 8013 - Posted: 8 Jun 2005, 12:07:55 UTC

I guess I am lucky, I get estimates from 11 to 13 hours. It looks like I get them done in about 4 too ... but these estimates are not that bad ...
ID: 8013 · Report as offensive     Reply Quote
Profile sysfried

Send message
Joined: 27 Sep 04
Posts: 282
Credit: 1,415,417
RAC: 0
Message 8015 - Posted: 8 Jun 2005, 12:35:24 UTC - in response to Message 8013.  

> I guess I am lucky, I get estimates from 11 to 13 hours. It looks like I get
> them done in about 4 too ... but these estimates are not that bad ...
>
my opteron does them in somewhat 2:40 h while estimates are 9:18 h

ID: 8015 · Report as offensive     Reply Quote

Message boards : Number crunching : Time to Completion


©2024 CERN