Poll: Completion of WU's

Section supporting all the various Climate Change projects: CPDN, CPDN Beta, Quake Catcher.
Post Reply
Robert Boon
Posts: 34
Joined: Fri Mar 17, 2006 12:00 am

Poll: Completion of WU's

Post by Robert Boon »

I am presently on my 5th attempt to complete a WU, the four previous attempts failing at 16%, 64%, 96% (sob), and 7%. What are other peoples experience with CPDN, BBC CCE and Seasonal Attribution.
DJH@GB-Ro
Posts: 525
Joined: Wed Mar 15, 2006 12:00 am

Post by DJH@GB-Ro »

I've had only one failure with BBC CCE. And I've yet to complete a full one, I'm too new!
UBT - BHCJackie
Posts: 315
Joined: Mon Mar 13, 2006 12:00 am

Re: Poll: Completion of WU's

Post by UBT - BHCJackie »

Robert Boon wrote:What are other peoples experience with CPDN, BBC CCE and Seasonal Attribution.
I'm on my 7th BBC WU in 6 weeks. I'm pretty sure I've not completed one but I don't know what the result id's mean so I can''t confirm my assumption.
UBT - Halifax-lad
Posts: 3790
Joined: Mon Mar 13, 2006 12:00 am

Post by UBT - Halifax-lad »

You won't have completed one yet Jackie if you have had so many in the last few weeks.

Also people who have that many failures are best off ditching the project as you will find it very hard to ever return a complete WU
UBT - BHCJackie
Posts: 315
Joined: Mon Mar 13, 2006 12:00 am

Post by UBT - BHCJackie »

UBT - Halifax--lad wrote:You won't have completed one yet Jackie if you have had so many in the last few weeks.

Also people who have that many failures are best off ditching the project as you will find it very hard to ever return a complete WU
IIRC the Mods on the BBC site said some Planet Earth's could go critical, ice age etc, within a short time so it was best to keep going.
This is why I'd really like to know what the result codes mean.

I must have done something right or I wouldn't have any credit, would I?
I've honestly thought of leaving the BBC experiment on more than one occasion. I like simple things like HashClash
UBT - Halifax-lad
Posts: 3790
Joined: Mon Mar 13, 2006 12:00 am

Post by UBT - Halifax-lad »

Everyone gets credit for a CPDN WU as it send a trickle of info to the servers everytime you reach a new time step, this awards you with credit.

So even with all your failed WU's you still get credit due to the trickle up messages
UBT - Tony
Posts: 8
Joined: Mon Mar 27, 2006 1:00 am

Post by UBT - Tony »

How do I find out how many WU's I have done?
I have credits but it means nothing to me :roll:
Cheers
Tony
Francis McDermott
Active UBT Contributor 15+ yrs
Posts: 73
Joined: Sat Mar 25, 2006 12:00 am
Location: Northamptonshire

Post by Francis McDermott »

I've had two CPDN WUs fail on me (only got to a few percent on each tho) and am on my third one. When this fails (I don't expect it to get to 100%) I'm moving the project onto a different machine which will concentrate more of its time on it.

I liked the sound of the BBC but would never have finished one this year!
Robert Boon
Posts: 34
Joined: Fri Mar 17, 2006 12:00 am

Post by Robert Boon »

Just had a sulphur model fail at 41% having spent over 1700 hours on it.  Acoording to the CPDN site I have crunched over the equivalent of 3.5 Hadsm3 WU but in reality I've NEVER had a WU complete successfully. :cry:
UBT - Timbo
UBT Forum Admin
Posts: 9673
Joined: Mon Mar 13, 2006 12:00 am
Location: NW Midlands
Contact:

Re: Poll: Completion of WU's

Post by UBT - Timbo »

Robert Boon wrote:I am presently on my 5th attempt to complete a WU, the four previous attempts failing at 16%, 64%, 96% (sob), and 7%.  What are other peoples experience with CPDN, BBC CCE and Seasonal Attribution.
I have had quite a few CPDN WU's but sadly, all but one failed early one, either due to hardware failures or "something else".

Had 1 CPDN WU, that almost completed but at some point near the very end it tripped up.

No one ever tells you why though - so it's either "one of those things" or something peculiar with either the WU or the PC crunching it.

But these WU's are pretty intensive on the hardware and it only needs a little error to screw the whole thing.

Mostly though I suspect the problem (if there is one) is linked to BOINC Manager switching between projects as I've had a few issues when a PC is linked to say two or three projects and then one switches over to another - I lost a SETI Beta WU today like that - BOINC switched from SETI SE to SETI Beta and then it came up at 100% completion, even though it was only about 20 hours through a 30 hours (expected) WU.

Maybe if you had only ran one project (and it was CPDN) then you'd be OK as no switching happens !!  :?:


I've not crunched a CPDN WU for ages due to these issues.

I've also lost some Seasonal WU and a BBC CCE (all with client computing error) - all due to "something"...!


I've actually got to the point now where I prefer crunching lots of short WU's as at least if one WU fails, you only lose an hour or so - SETI (v4.18 - not the new SE !!), Leiden is a good candidate right now, as is Einstein with an opti-client.

PrimeGrid and Predictor used to have short length WU's


regards,

Tim
UBT - Halifax-lad
Posts: 3790
Joined: Mon Mar 13, 2006 12:00 am

Post by UBT - Halifax-lad »

Switching works here fine between 20+ projects
UBT - Timbo
UBT Forum Admin
Posts: 9673
Joined: Mon Mar 13, 2006 12:00 am
Location: NW Midlands
Contact:

Post by UBT - Timbo »

UBT - Halifax--lad wrote:Switching works here fine between 20+ projects
The one's I found more prone are:

SETI Beta
CPDN
Rosetta (v4.x)

Most others are OK - I've now actually set the BOINC Preference to switch applications now (edit: ) LESS OFTEN (to every 120 minutes and in time may up this to 240 minutes) as I'm doing less crunching on every project and am manually switching projects "as and when" the mood takes me - esp. for things like "crunches" or when we need a bit of a push...!

And mostly I now only have two active projects per PC - so if one runs out the other can take over - and with a 1.5 day cache, this seems to work well - very rare to have a PC run totally dry...!

regards,

Tim
Last edited by UBT - Timbo on Fri May 12, 2006 6:17 am, edited 1 time in total.
Robert Boon
Posts: 34
Joined: Fri Mar 17, 2006 12:00 am

Post by Robert Boon »

When the CPDN WU crashed I had all both of my other projects (Seti and Rosetta) set to receive no new work and had all Seti and Rosetta WU's suspended so the problem, whatever it is, is certainly not caused by switching between projects.

It may be relevant that if the CPU benchmark tests run whilst I am running CPDN I always find that the CPDN project fails to restart once the test finishes and I have to switch between projects in order to restart CPDN.
RodEllery
Posts: 489
Joined: Fri Mar 24, 2006 12:00 am

Post by RodEllery »

CPDN seems 100% reliable here, Only completed 2 WUs but current 2 WUs have run for 1129 and 179 hrs respectively and have 733 and 1927 hrs to go.

Only time I had a problem was recovering from a motherboard failure.

--
Rod
MikeMarsUK
BOINCSynergy team member
Posts: 110
Joined: Sat Aug 12, 2006 1:00 am

Post by MikeMarsUK »

A few WU failures, but every time it was me that caused it in some way (i.e., killing a SAP task at 85% in task manager when trying to back it up, etc).

To run the climate models successfully, the PC needs to be Prime95-stable for at least 24 hours (in some ways CPDN is the harshest test of PC stability there is).  It's also sensitive to problems with graphics drivers etc.  

Post problems on http://www.climateprediction.net/board/index.php and we'll try to help, a 'messages' tab dump is always useful too.
danieljones2006
Posts: 1
Joined: Sat May 01, 2010 1:00 am

Post by danieljones2006 »

Minority of WU's complete - mostly failures is what the status with me!!!
UBT - Rick Horn
Posts: 17206
Joined: Sat May 06, 2006 1:00 am

Post by UBT - Rick Horn »

I would look on the message board of the project you are using, and see if there is a general problem or not.
Kevy
Posts: 11
Joined: Sat Mar 19, 2011 12:00 am

Post by Kevy »

I've had a couple of errors, and another 2 'error while downloading', but otherwise fine

Image
catchercradle
Posts: 5
Joined: Thu Dec 16, 2010 12:00 am

Post by catchercradle »

Something that isn't addressed here is that the non-completion of a work unit in CPDN does not always imply a failure. Part of the work is to see which tasks produce possible climates. If they error with a -ve theta it means an impossible climate e.g. a negative pressure. This is still useful information.

There has been a lot of discussion about this on the CPDN boards. Most of the recent units are long over 800 hours on this computer and on my dual core atom over 3,000 hours! This gives a lot of chances for a model to fail - if an anti-virus program puts a lock on a file when boinc wants to write to it this can cause it to fail for example. Power failures are another common cause of models falling over. Don't know whether any of this will change after March when it is planned to upgrade the boinc server software or if it is better with with the BOINC7.x - I seem to be getting fewer failures now than I have in the past.

Dave
Post Reply