The problem with deadline for me is that no matter how I will set it, it is always to short for the downloaded units to be crunched - and then as I understand, no credit applies. Additionally, if today's outage will last till those 10 hours, I will lost the credit for another 3 wu's which have deadline before then.

The problem is not directly with the deadline itself but most likely the scheduler/client or whatever that calculates how much work to send to us. It sends too much work. If I set the connection interval to 14 days, it downloads 16 days of work but sets the deadline for 14 days, if I set 7 days, it downloads 8 days of work (approx) and sets the deadline for 7 days.

I have been explained that this will be fixed in future by fixing the calculation method or whatever. But..

Wouldn't it be simplier and better to make a temporary solution just until someone will start fixing all issues here...?

The temporary solution is quite simple: increase the deadline from "1x connection interval" to something around "1.3x connection interval". That will likely save a lot of trouble...

If you are not finishing WU's until nearly past the deadline... Then the problem is at your end, not at the server end. (The deadline is always two weeks, it has nothing to do with your connection interval.)

Either your computer is too slow, or you are not running S@H B for enough hours in the day, or it's priority in your BOINC setup is too low, or you are running too many projects, or you are not connecting often enough.

> If you are not finishing WU's until nearly past the deadline... Then the
> problem is at your end, not at the server end. (The deadline is always two
> weeks, it has nothing to do with your connection interval.)
>
> Either your computer is too slow, or you are not running S@H B for enough
> hours in the day, or it's priority in your BOINC setup is too low, or you are
> running too many projects, or you are not connecting often enough.
>
>

I have a IBM/Cyrix 166 on my network running (BOINC/S@H) 24/7, and it averages about 100 hours/WU (well within the fortnight allotted), so it has to be a personal setting issue (adjust it so it runs more often).

> If you are not finishing WU's until nearly past the deadline... Then the
> problem is at your end, not at the server end. (The deadline is always two
> weeks, it has nothing to do with your connection interval.)
>
> Either your computer is too slow,

Athlon 1333 MHz

> or you are not running S@H B for enough hours in the day,

24 h/day, "run always"

> or it's priority in your BOINC setup is too low,

priority is default

> or you are running too many projects

only SETI@home

> or you are not connecting often enough.

I have set the "Connect to network about every ..." for 7 days and reset the project. This is because I want to have enough units cashed in case of either my connection or the S@h servers failure. After reset, client has downloaded a work that had a deadline 7 days. Yet, two units has been completed to late and another one was completed in time but the server was out and it was up after the deadline.

For all other units, client was uploading the result as soon as finished crunching, and as far as I understand this client it reported the "ready to report" results while I was clicking "update".

Therefore, I list things that are IMO wrong generally with the BOINC platform:

a) the deadline estimation concept is wrong, because
- the algorithm that calculates the deadline is not perfect, and it may return a slightly too short or too long time
- the platform does not allow certain time tollerance (while it should based on the algorithm's inaccuracy)

b) the client itself does not care in any way about the deadline
- it calculates the workunit even if it is past the deadline
- it tries to upload the crunched but deadlined workunit
which is a waste of calculating power both for me and for the project which rejects the deadlined workunits

Then it comes to the conclusion, that each time I have reached a deadline and have the work remaining that is over deadline, I must reset the project.

Unfortunately, in the same time I have some other workunits that have been downloaded by the client. Currently, I have reset the project and therefore were forced to discard non-deadlined workunits too.

If someone else downloaded the same workunits and crunched them, he is in higher risk that he will not get any credit for them since my copy of the workunits will never be calculated and reported for the validation.


There is one other solution: a feature in client that allows you to manually select and discard workunits that are past the deadline.

Whatever the solutions are, one must be found because in current state, the project wastes the computing power.

> > If you are not finishing WU's until nearly past the deadline... Then
> the
> > problem is at your end, not at the server end. (The deadline is always
> two
> > weeks, it has nothing to do with your connection interval.)
> >
> > Either your computer is too slow,
>
> Athlon 1333 MHz
>
> > or you are not running S@H B for enough hours in the day,
>
> 24 h/day, "run always"
>
> > or it's priority in your BOINC setup is too low,
>
> priority is default
>
> > or you are running too many projects
>
> only SETI@home
>
> > or you are not connecting often enough.
>
> I have set the "Connect to network about every ..." for 7 days and reset the
> project. This is because I want to have enough units cashed in case of either
> my connection or the S@h servers failure. After reset, client has downloaded a
> work that had a deadline 7 days. Yet, two units has been completed to late
> and another one was completed in time but the server was out and it was up
> after the deadline.
>
> For all other units, client was uploading the result as soon as finished
> crunching, and as far as I understand this client it reported the "ready to
> report" results while I was clicking "update".
>
> Therefore, I list things that are IMO wrong generally with the BOINC
> platform:
>
> a) the deadline estimation concept is wrong, because
> - the algorithm that calculates the deadline is not perfect, and it may return
> a slightly too short or too long time
> - the platform does not allow certain time tollerance (while it should based
> on the algorithm's inaccuracy)
>
> b) the client itself does not care in any way about the deadline
> - it calculates the workunit even if it is past the deadline
> - it tries to upload the crunched but deadlined workunit
> which is a waste of calculating power both for me and for the project which
> rejects the deadlined workunits
>
> Then it comes to the conclusion, that each time I have reached a deadline and
> have the work remaining that is over deadline, I must reset the project.
>
> Unfortunately, in the same time I have some other workunits that have been
> downloaded by the client. Currently, I have reset the project and therefore
> were forced to discard non-deadlined workunits too.
>
> If someone else downloaded the same workunits and crunched them, he is in
> higher risk that he will not get any credit for them since my copy of the
> workunits will never be calculated and reported for the validation.
>
>
> There is one other solution: a feature in client that allows you to manually
> select and discard workunits that are past the deadline.
>
> Whatever the solutions are, one must be found because in current state, the
> project wastes the computing power.
>
>

Discarding WU's is a waste of bandwith, since uploading/downloading consume bandwith. I dont see a reason to change the C C again as things start to stabilise. Adjust your cache to meet the dfeadline. I would never download more than what my machine can crunch.

> Discarding WU's is a waste of bandwith, since uploading/downloading consume
> bandwith. I dont see a reason to change the C C again as things start to
> stabilise. Adjust your cache to meet the dfeadline. I would never download
> more than what my machine can crunch.

I well know that discarding is a waste of many resources. But I have no choice since BOINC is stubborn to set the deadline inaccuratly. I download tens of packages for seven days, but it always gives me work for 8 days and the dealine for 7 days.

Until solved, I have no choice.

> > Discarding WU's is a waste of bandwith, since uploading/downloading
> consume
> > bandwith. I dont see a reason to change the C C again as things start to
> > stabilise. Adjust your cache to meet the dfeadline. I would never
> download
> > more than what my machine can crunch.
>
> I well know that discarding is a waste of many resources. But I have no choice
> since BOINC is stubborn to set the deadline inaccuratly. I download tens of
> packages for seven days, but it always gives me work for 8 days and the
> dealine for 7 days.
>
> Until solved, I have no choice.
>
>

Have you tried 1 day or 2 days or 0.2 days, or just stop processing if you keep discarding WUWUs, it is not fair to the others who process.....



My name is Pascal and this message has no meaning, but still has my approval...

It is 10 oclock, do you know what your WUWUs are doing tonight...

> Have you tried 1 day or 2 days or 0.2 days, or just stop processing if you
> keep discarding WUWUs, it is not fair to the others who process.....

But like I said I need the cache for several days in case of server problems.

I well know it is unfair but I am not guilty of that. It is same unfair for all of us, bot you and me and others.

> I download tens of
> packages for seven days, but it always gives me work for 8 days and the
> dealine for 7 days.
>
> Until solved, I have no choice.
>
>

The deadline is ALWAYS supposed to be set for 14 days regardless of how bad BOINC's benchmarks misinterpret crunching capability of a host. Remember this deadline is set from the time the scheduler "releases" those work units to be transfered to the host, I don't think the deadline gets set when they are finished downloading to the host. For example I have a WU that finished download to the host at 14:30:28 on 10/05/04, but the deadline is set at 14:22:36 on 10/19/04. It was the last to download of a group of about 10 WU's with deadline times within 1 second of each other.

If you have work units with deadlines that are not 14 days from the day the scheduler assigned them to your host, please post some lines from your messages tab showing when they finished downloading and the info from your work tab showing the same WU and the deadline date.

Maybe we can help figure it out.
example:

SETI@home - 2004-10-05 14:30:28 - Finished download of 26ap04aa.7796.14048.384654.194

Seti@home setiathome 4.05 26ap04aa.7796.14048.384654.194_4 --- 0.00% 5:44:18 2004-10-19 14:22:36 Ready to run

Please stop resetting the project, and adjust your cache lower.

All WUs (currently) have a deadline of 14 days! Adjusting your cache to 7 days does not give you WUs with a 7 day deadline!

In your words:
"I download tens of packages for seven days, but it always gives me work for 8 days and the dealine for 7 days."

I suggest you look closely at the download dates and deadlines assigned, because this is not how the system works...

By resetting the project every time you blow by a deadline because you have your cache set to large to complete all the work by the time it is due, you are delaying many other peoples credits. This is not very cool.

Please just set your cache smaller, and leave it. Sure you may run out of work sometimes due to server outages, but the countless people you are quietly screwing by your thoughtless resets will thank you.

Alternatively, attach to another project. Chances are good that one will be giving work while the other is down.

Seti wu's have 14 day deadline from when it is downloaded to your computer.
Set your connect ever xxx days to x days.
Not only affecting credit, you are screwing up the scheduler server.
Fred BOINC Alpha & BOINC Beta Tester

> By resetting the project every time you blow by a deadline because you have
> your cache set to large to complete all the work by the time it is due, you
> are delaying many other peoples credits. This is not very cool.

By not resetting, my BOINC calculates workunits which I will not recieve any credit for. That is not very cool too.

> Alternatively, attach to another project. Chances are good that one will be
> giving work while the other is down.

Not an option


Generally I am waiting now for the reset of the project (I have firstly reset the project, then some errors occured, reinstalled it, but the error is still there, cannot dload anything now, check http://setiweb.ssl.berkeley.edu/forum_thread.php?id=5215 ) and then I will be able to tell you how the thing is going on, is it still deadlining wrongly or was it a problem with installation.

BTW.. since there is sth like half a day time difference between US and me, is it possible to be a cause for these additional units assingned to me (those over deadline)?

> > By resetting the project every time you blow by a deadline because you
> have
> > your cache set to large to complete all the work by the time it is due,
> you
> > are delaying many other peoples credits. This is not very cool.
>
> By not resetting, my BOINC calculates workunits which I will not recieve any
> credit for. That is not very cool too.
>
> > Alternatively, attach to another project. Chances are good that one will
> be
> > giving work while the other is down.
>
> Not an option
>
>
> Generally I am waiting now for the reset of the project (I have firstly reset
> the project, then some errors occured, reinstalled it, but the error is still
> there, cannot dload anything now, check
> http://setiweb.ssl.berkeley.edu/forum_thread.php?id=5215 ) and then I will be
> able to tell you how the thing is going on, is it still deadlining wrongly or
> was it a problem with installation.
>
> BTW.. since there is sth like half a day time difference between US and me, is
> it possible to be a cause for these additional units assingned to me (those
> over deadline)?
>
>
IF YOU SET YOUR CONNECT EVERY DAY TO A REASONABLE TIME YOU WOULD NOT HAVE TO RESET!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
Fred BOINC Alpha & BOINC Beta Tester

> IF YOU SET YOUR CONNECT EVERY DAY TO A REASONABLE TIME YOU WOULD NOT HAVE TO
> RESET!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

It seems that no matter how many people tell him he's doing it wrong and that he doesn't actually understand how the system works, he's just going to ignore us and continue screwing up the project, and delaying credits for the unlucky folks also being sent the same WUs. Then blame what he's doing on the system... :-( :-( :-(

Kind of reminds me of JQ...

> Kind of reminds me of JQ...

That name keeps coming up. Tell me more about JQ?

Reading Threads like this, helps to explain why 90% of the work I've gotten today ends with a 3 or higher. One work unit had been sent out 7 times.

At least some users will be getting their credit soon. My cache is only set for 2 days and I've been lucky with Boinc. I haven't lost a work unit yet. :)

Just a suggestion, but let's not harsh this guy too much. He is posting a question because something's going on with his work units and his machine that he doesn't quite understand. Again, just a suggestion. Flaming him may not help him learn the ropes.
Johann, consider taking the suggestion of the person (Venus and Mars)that recommended you post some of your message tabs so folks can take a look at what's happening on your computer.
Easy does it.
Robert

Johann "onn" Bitsoenn,

Okay, let's try this one more time. The "Connect every" value has NOTHING TO DO with the period between the time you download the WU and when it "expires." This duration is a constant 14 days. PERIOD.

DO NOT SET YOUR "CONNECT EVERY" VALUE TO MORE THAN FIVE (5) DAYS. FIVE (5) IS THE MAXIMUM recommended TIME INTERVAL. Unless they've change it very recently, a setting of D days will give you somewhere between D days and 2D days of work, based on your benchmarks.

For you, I might recommend a "Connect Every" value of 3, giving you between 3 and 6 days of work. This will cover any short server outages. Set the "Connect Every" value and leave it alone. Don't screw with it. This should also give you few enough WUs to complete well within the deadline.

BTW, I understand the SETI v4.05 science app gives calculation times that are significantly longer than v4.03 did. I still have several 4.03 WUs to process before I start on the 4.05 WUs.

How much RAM do you have installed? The cheapest way to speed up a computer is to add more RAM. What other applications are you running? What do you have running in the System Tray? Are you leaving BOINC_gui open? What tab are you leaving the display on? I recommend selecting either the "Projects" or "Messages" tab. I also recommend hiding the app when you don't "have" to have it open, and select the "Projects" tab before you hide it.

When you do a reset, you tend to annoy (also know as "piss off") a great number of people because you make the WUs you lose into "orphans," and people, who have been waiting for your results so they can get credit for the work THEY have done, have to wait even longer. They now have to wait until the WU is resent, processed, and the results returned by yet another person. This is not how to win friends and influence people. Then, when they start whinning about not getting credit, I have to read through all of that "stuff."

Stephen
<P>"I want to go dancing on the moon, I want to frolic in zero gravity!....", and now, I might be able to go someday! Thanks, SpaceShipOne and crew!<BR><a> [/url]

> Wouldn't it be simplier and better to make a temporary solution just until
> someone will start fixing all issues here...?
>
> The temporary solution is quite simple: increase the deadline from "1x
> connection interval" to something around "1.3x connection interval". That will
> likely save a lot of trouble...

Johann,

I know it is "un-cool" to suggest this ... :)

But, click on my signature and do a little reading. That may help you with your problem. For you, I would suggest looking up the words relating to credit, deadlines, etc. in the Glossary.

In addition, there are pages that "mirror" the SETI@Home web site so that you can read up on the settings and then jump directly to the page on the SETI@Home web site to make changes. So, navigate just like you would to the pages that describe the "Your Account" web page. Then read up on the line by line settings.

If any of that material is confusing, or you don't understand the explanation, drop me an e-mail and I will try again.

One more general tip, the pages are generic and you can click on the buttons to look at the alternative projects and how their pages look....
<p>
For BOINC Documentation: Click Me!

I would like to see a Screen Shot of this hypothetical problem.




My name is Pascal and this message has no meaning, but still has my approval...

It is 10 oclock, do you know what your WUWUs are doing tonight...