JM7...I respect your knowledge and diligence...what is your take on the WU shortage?

I don't know when the WU shortage will be over. I have to believe the news that there were an unxpectly large number of machines that crossed over to S@H, and the developers are working on fixes. I am betting that they are not having muchof a weekend.

I have decided to share my crunching power among projects. I am getting WUs from predictor while S@H is having difficulty. Of course the choice is always yours. If you wanted to get from S@H except when S@H is down, you can set your preferences to be .001 for Predictor and 1000 for S@H.

Couple of points (inline)

> I don't know when the WU shortage will be over. I have to believe the news
> that there were an unxpectly large number of machines that crossed over to
> S@H, and the developers are working on fixes. I am betting that they are not
> having muchof a weekend.

The Beta had about 20k registered users, and last that I could check, BOINC had about 30k. That doesn't seem like an "unxpectly large number of machines". Sorry.. they didn't plan the switchover. Having 50% more users go live than those that were in the beta doesn't strike me as unexpected! :(

>
> I have decided to share my crunching power among projects. I am getting WUs
> from predictor while S@H is having difficulty. Of course the choice is always
> yours. If you wanted to get from S@H except when S@H is down, you can set
> your preferences to be .001 for Predictor and 1000 for S@H.

Doesn't a "predictor" WU take on the order of days to run(if not weeks), even on a very fast computer? Wouldn't you potentially exculde your system from Seti for quite some time, if a seti/boinc outage problem gets resolved just an hour or so? What happens to those seti WU's that are "ready to report", but will sit in limbo for up to a couple weeks? (the seti/boinc server won't be contacted for a long time) Yes, up can force an "update" sooner, but most user's won't do that.

WOODY...

> Doesn't a "predictor" WU take on the order of days to run(if not weeks), even
> on a very fast computer? Wouldn't you potentially exculde your system from
> Seti for quite some time, if a seti/boinc outage problem gets resolved just an
> hour or so? What happens to those seti WU's that are "ready to report", but
> will sit in limbo for up to a couple weeks? (the seti/boinc server won't be
> contacted for a long time) Yes, up can force an "update" sooner, but most
> user's won't do that.
>

I seem to recall some posts regarding 'Trickle Messages'...

Please do your homework!

> WOODY...
>
> > Doesn't a "predictor" WU take on the order of days to run(if not weeks),
> even
> > on a very fast computer? Wouldn't you potentially exculde your system
> from
> > Seti for quite some time, if a seti/boinc outage problem gets resolved
> just an
> > hour or so? What happens to those seti WU's that are "ready to report",
> but
> > will sit in limbo for up to a couple weeks? (the seti/boinc server won't
> be
> > contacted for a long time) Yes, up can force an "update" sooner, but
> most
> > user's won't do that.
> >
>
> I seem to recall some posts regarding 'Trickle Messages'...
>
> Please do your homework!
>
Both of you do your homework. I should have specified more clearly. There is a Predictor program that predicts the folding pattern of a protien, it can be found at http://predictor.scripps.edu/. It is not the same as climate prediction.net. Predictor WUs on average are taking about 2/3 of the time that a S@H WU takes. A CPDN WU is indeed expected to take a loooong time, estimate 3 weeks on a 3GHz machine is a rumor that I heard.

Ah yes, the buzzard is back....

Anyway folks, sorry for the poster with the beady eyes.

Anyway, back to to question. If a "predictor" WU takes days to process, doesn't that limit the BOINC communications back to seti for RTR WU's?

Hey, I will say, I'm not a fan of BOINC. I was a beta tester for 7 months and have been with SETI almost since day 1. BOINC would be a "brave new world" even if things worked, but right now, I see the same bugs that have been there for 6 months, and the launch was VERY premature! Seti 1 in birthing had a few problems (i was a cruncher back then), but Seti 2 isn't a birthing - it's just "moving to a new house". Seems they moved to a two bedroom/1 bath, and now find they need a 4 bedroom/2 bath after counting heads!

Some buzzards in cheerleader garb think all is cool as they were allowed to beta test, as it's their biggest thrill since being made library monitor! Maybe the parents will allow them to drive one day!

If I complained, without trying to use it, I'd just be a buzzard, and peck at meat I thought was dead! (Kind of like a buzzard we know).

"I'm not dead yet!" (MP + the Holy Grail)

> Anyway, back to to question. If a "predictor" WU takes days to process,
> doesn't that limit the BOINC communications back to seti for RTR WU's?

Well, since you ignored Johns comment, I'll make it as well. Maybe that will help. "Predictor" WUs do not take days to process. They are much quicker than S@H WUs.

> > Anyway, back to to question. If a "predictor" WU takes days to process,
> > doesn't that limit the BOINC communications back to seti for RTR WU's?
>
> Well, since you ignored Johns comment, I'll make it as well. Maybe that will
> help. "Predictor" WUs do not take days to process. They are much
> quicker than S@H WUs.

I've not run "predictor" itself, but from what I've read, it does take days-weeks. I'm sorry if I got it wrong, but I had this buzzard trying to peck at me!

What about climate-predictor? Is that the one that takes days-weeks? Or am I thinking about the project that models protiens? Anyway, replace "predictor" with "another boinc project with WU's that takes days to run a single WU", and the question still stands! Doesn't running another BOINC project WU with LONG times impact Seti as RTR WU's don't get reported?

(don't address the message, attack the messanger! A true Boinc cheeleader!)

(BTW, I didn't see VII's response until after my response to the Buzzard!! Gosh, you guy's are getting testy these days!)

WOODY...

> Ah yes, the buzzard is back....

Ummmm... there is a BIG difference between a buzzard and a crow... maybe if you had some 'real life expereince' you would know that...

>I've not run "predictor" itself, but from what I've read, it does take days->weeks. I'm sorry if I got it wrong, but I had this buzzard trying to peck at >me!

OMGWAFDA** (do you really think that WOODY actually Reads?)...

>>>> lil' bro... DO YOUR HOMEWORK... before you post...

>It is not the same as climate prediction.net. Predictor WUs on average are
>>taking about 2/3 of the time
> that a S@H WU takes. A CPDN WU is indeed expected to take a loooong time,
> estimate 3 weeks on a 3GHz machine is a rumor that I heard.

VII - that's the one I was thinking (3w on a 3gh). Using the non-Boinc version, I got to .25% in the first set in a day!

But... The question still remains.. If I run climate-prediction with 3week WU's, won't that screw up others that are waiting for my SETI WU's to move past RTR? Best case is other's waiting for pending to move to granted will be delayed. Worst case (to me) is the RTR will miss the deadline and I'll get zip, and other's may need to wait 2 weeks past the deadline for the third WU!

>
> >It is not the same as climate prediction.net. Predictor WUs on average
> are
> >>taking about 2/3 of the time
> > that a S@H WU takes. A CPDN WU is indeed expected to take a loooong
> time,
> > estimate 3 weeks on a 3GHz machine is a rumor that I heard.
>
> VII - that's the one I was thinking (3w on a 3gh). Using the non-Boinc
> version, I got to .25% in the first set in a day!
>
> But... The question still remains.. If I run climate-prediction with 3week
> WU's, won't that screw up others that are waiting for my SETI WU's to move
> past RTR? Best case is other's waiting for pending to move to granted will be
> delayed. Worst case (to me) is the RTR will miss the deadline and I'll get
> zip, and other's may need to wait 2 weeks past the deadline for the third WU!
>
>
You don't need to worry about the workunit being returned late. BOINC will report it about 24 hours before it is due.

John Keck
BOINCing since 2002/12/08

> But... The question still remains.. If I run climate-prediction with 3week
> WU's, won't that screw up others that are waiting for my SETI WU's to move
> past RTR? Best case is other's waiting for pending to move to granted will be
> delayed. Worst case (to me) is the RTR will miss the deadline and I'll get
> zip, and other's may need to wait 2 weeks past the deadline for the third WU!

I don't think it will have any effect. I believe BOINC will go by expiration dates in an instance such as this.

Edit: Ah, John Keck beat me to it. :)

>
> >It is not the same as climate prediction.net. Predictor WUs on average
> are
> >>taking about 2/3 of the time
> > that a S@H WU takes. A CPDN WU is indeed expected to take a loooong
> time,
> > estimate 3 weeks on a 3GHz machine is a rumor that I heard.
>
> VII - that's the one I was thinking (3w on a 3gh). Using the non-Boinc
> version, I got to .25% in the first set in a day!
>
> But... The question still remains.. If I run climate-prediction with 3week
> WU's, won't that screw up others that are waiting for my SETI WU's to move
> past RTR? Best case is other's waiting for pending to move to granted will be
> delayed. Worst case (to me) is the RTR will miss the deadline and I'll get
> zip, and other's may need to wait 2 weeks past the deadline for the third WU!
>
No worse than some dialup user (they still exist) that queues 14 days of work, because he or she only connects once per week. Or someone with a really slow computer (I have one of these) that takes nearly a week to crunch a WU. The only way you get hurt is if there is a server outage that takes the whole day that you are trying to beat deadline.

> I don't think it will have any effect. I believe BOINC will go by expiration
> dates in an instance such as this.

Actually, it does have an effect, but not on the science!

Let's say that the first two crunchers got their results back in a day, and they're sitting with pending credit on the WU. The third result might not be back for another 9 days! If all three are OK, credits are granted, but if not, it'll be another 10 days!

Look at Seti 1 - here's a WHOLE bunch of folks that watch credits/performance on a dayly basis! This one case could result in "credits" being delayed 20-30 days!

Prime example of BOINC not taking into considersation the end users! There are many that watch their numbers, and if their numbers only increase after two weeks based on the actions/incations of a user they've never heard of, it won't sit well!

Give the users "WU's sent/WU's returned" numbers, and the crunchers will be happy, as they won't get P.O.'ed at someone halfway around the world that didn't get their WU done in time(or it had an error on their system, for which they they had no control!)

> Actually, it does have an effect, but not on the science!
>
> Let's say that the first two crunchers got their results back in a day, and
> they're sitting with pending credit on the WU. The third result might not be
> back for another 9 days! If all three are OK, credits are granted, but if
> not, it'll be another 10 days!

Weren't you asking how projects with long WU times such as climate prediction would effect the credit outcome???

What you described is exactly what you can expect if you don't manually return your results. I suspect many will just wait for automatic retrieval. If by the worst case scenario this goes by expiration date, (assuming they haven't picked up work recently) you will see credit turnaround times like you mention.

Climate prediction doesn't aggravate this condition in the slightest.

> Prime example of BOINC not taking into considersation the end users! There
> are many that watch their numbers, and if their numbers only increase after
> two weeks based on the actions/incations of a user they've never heard of, it
> won't sit well!
>
> Give the users "WU's sent/WU's returned" numbers, and the crunchers will be
> happy, as they won't get P.O.'ed at someone halfway around the world that
> didn't get their WU done in time(or it had an error on their system, for which
> they they had no control!)

Unfortunately this statistic lead to rampant cheating to get ahead on the numbers, never mind if it cheated the science. BOINC is now insisting that the science comes first, and the numbers second. Remember that everyone has the same average delay in getting credit, so the leader boards are a valid indication of who has done the most work. Hopefully, the system is robust enough that cheating will be difficult enough that the cheaters will either not cheat, or leave. Either is fine. This leaves us with the core that does not cheat, and I hope the delay in granting credit will not turn tooo many of these away. Please note that overclocking your machine to such an extent that it cannot do floating point arithmetic with enough accuracy and stability to get the science done is included as a cheat in my book. Maybe not quite as intentional as returning pre-crunched results, but it has the same effect on the science.

> > Prime example of BOINC not taking into considersation the end users!
> There
> > are many that watch their numbers, and if their numbers only increase
> after
> > two weeks based on the actions/incations of a user they've never heard
> of, it
> > won't sit well!
> >
> > Give the users "WU's sent/WU's returned" numbers, and the crunchers will
> be
> > happy, as they won't get P.O.'ed at someone halfway around the world
> that
> > didn't get their WU done in time(or it had an error on their system, for
> which
> > they they had no control!)
>
One thing you seem to be missing is that after a few weeks the pending credit tends to stablize (current difficulties excluded). Most users will see their credit rising every time they connect because some will have been validated from the pool.

John Keck
BOINCing since 2002/12/08