What gives here, and when will I get credit?

http://setiweb.ssl.berkeley.edu/workunit.php?wuid=5661638

Why isn't credit being allocated with 3 results in already, and when will I get it?

---

All 3 Pending WU's show "Checked, but no consensus yet", so the Validator didn't like something about them I would take it and the WU was sent out again for another person to do to see if it matched the other returned results better ...

If there is an answer I'd appreciate the same for work units 347949,162257,909934 all of which show three sucessful results but still showing as pending in my account. And even stranger units 5796714,877653,8909744,8909744 which all show 4 sucessful results and nothing credited and showing pending still.


> What gives here, and when will I get credit?
>
> http://setiweb.ssl.berkeley.edu/workunit.php?wuid=5661638
>
> Why isn't credit being allocated with 3 results in already, and when will I
> get it?
>

---

> If there is an answer I'd appreciate the same for work units
> 347949,162257,909934 all of which show three sucessful results but still
> showing as pending in my account.

"Too many total results", in other words these wu have errored-out and will not give credit.

> And even stranger units
> 5796714

The same as in 1st. post in this thread, "No consensus yet". These wu is re-issued to either a consensus is reached, or errors out. It means, the validator didn't find a pair of results to be "nearly similar".

> 877653

This wu should have been re-surrected then the fix for re-issuing inconclusive wu was added beginning of December. But since from August, it's possible got permanently corrupted in one of the crashes...

> 8909744,8909744 which all show 4 sucessful results and nothing
> credited and showing pending still.
>

Unknown wu..., sure it's not a typing-error, especially since repeated the same wu twice, and currently only at around 7M wu?

At least in run time the first 2, my Athlon XP 1900+ and the other PC are close. It taking slightly less time makes sense, as it was a dual Pentium 4 2.8 GHz... Not sure about Apple's off the top of my head so can't comment. People not returning results (was addressed in other threads with people setting their queues too high and stuff).

Are we essentially screwed for getting credit, will we ever see it, what's going on?

---

> At least in run time the first 2, my Athlon XP 1900+ and the other PC are
> close. It taking slightly less time makes sense, as it was a dual Pentium 4
> 2.8 GHz... Not sure about Apple's off the top of my head so can't comment.
> People not returning results (was addressed in other threads with people
> setting their queues too high and stuff).

The run time is not relevant to validation. What matters is how similar the actual calculations from the unit are from machine to machine. Thus, it is not unusual to have units done by machines with very different run times that all receive credit.

> Are we essentially screwed for getting credit, will we ever see it, what's
> going on?

On the "errored out" units (those sent to the limit of number of hosts), you will never see credits as these units will never be resent.

> At least in run time the first 2, my Athlon XP 1900+ and the other PC are
> close. It taking slightly less time makes sense, as it was a dual Pentium 4
> 2.8 GHz... Not sure about Apple's off the top of my head so can't comment.
> People not returning results (was addressed in other threads with people
> setting their queues too high and stuff).
>
> Are we essentially screwed for getting credit, will we ever see it, what's
> going on?
>

Your wu have been re-distributed, so you can still get credit for this.

Looking more closely on the 3 computers, Amd, Intel & Apple. The powermac is mostly returning garbage and only on a very limited supply of his results is validation passed, so it's not unreasonable to expect his result isn't comparable with either of the 2 others for your wu.

Both the Amd & Intel seems to be performing correctly, but differences in cpu & os can sometimes gives so large difference that fails validation against eachother.
But, another p4 have now got this wu to crunch, and most likely this result then returned will be "similar enough" so wu now validated, and these 3 will get credit.

The powermac on the other hand very likely will fail validation for this wu also...

> > At least in run time the first 2, my Athlon XP 1900+ and the other PC
> are
> > close. It taking slightly less time makes sense, as it was a dual
> Pentium 4
> > 2.8 GHz... Not sure about Apple's off the top of my head so can't
> comment.
> > People not returning results (was addressed in other threads with people
> > setting their queues too high and stuff).
>
> The run time is not relevant to validation. What matters is how similar the
> actual calculations from the unit are from machine to machine. Thus, it is
> not unusual to have units done by machines with very different run times that
> all receive credit.
>

There is some relevence in a round about way. It indicates how long a given comp (in relation to each other) took to complete the given work unit (aka a given amount of work).

When one looks at performance benchmarks, pretty much any which can be used on a CPU, video card, etc...it has a given job to complete, and it measures how long it takes each competitive product to complete the work given. The use of standardized tests is how many review sites compare Intel vs. AMD's lattest gen processors, nVidia vs. ATI's lattest is compared, etc.

This said, it would give one reason to scratch their head if a Pentium 4 3 GHz took substantially longer to complete a given unit of work, then a 300 MHz Celeron, given what we know about the relative performance differences between these products. True to form with this, the computers that usually claim less credit then mine are either high GHz Pentium 4s (like P4 3.0as), and I'd expect Athlon 64s. Usually Celerons, Pentiums (have seen a couple) take a lot longer to complete then my AXP 1900+

Though it's differences in results that do matter, oddities in the time to completion can be indication of something. This is what I was taking into account, though I'm not familiar enough with Macs and how they stack up performance wise (in benchmarks for instance) to know if his was out of whack...

---

> Your wu have been re-distributed, so you can still get credit for this.
>
> Looking more closely on the 3 computers, Amd, Intel & Apple. The powermac
> is mostly returning garbage and only on a very limited supply of his results
> is validation passed, so it's not unreasonable to expect his result isn't
> comparable with either of the 2 others for your wu.

Yeah, it's just been redistributed. Thanks for the heads up. Hopefully all will turn out well for both this work unit, and the rest of us who processed it/will...

I wonder if they have a means to notify the user of that Mac concerning a problem with his results... Not sure what's going on, but obviously if he wants to contribute to the project he'll need to address his PC.

Anyhow, thx again...

---

Sorry was a misstype should be 890974 and 802411 both of which show 4 results all with no errors

> > If there is an answer I'd appreciate the same for work units
> > 347949,162257,909934 all of which show three sucessful results but still
> > showing as pending in my account.
>
> "Too many total results", in other words these wu have errored-out and will
> not give credit.
>
> > And even stranger units
> > 5796714
>
> The same as in 1st. post in this thread, "No consensus yet". These wu is
> re-issued to either a consensus is reached, or errors out. It means, the
> validator didn't find a pair of results to be "nearly similar".
>
> > 877653
>
> This wu should have been re-surrected then the fix for re-issuing inconclusive
> wu was added beginning of December. But since from August, it's possible got
> permanently corrupted in one of the crashes...
>
> > 8909744,8909744 which all show 4 sucessful results and nothing
> > credited and showing pending still.
> >
>
> Unknown wu..., sure it's not a typing-error, especially since repeated the
> same wu twice, and currently only at around 7M wu?
>

---

> Sorry was a misstype should be 890974 and 802411 both of which show 4 results
> all with no errors
>
> > > And even stranger units
> > > 5796714
> >
> > The same as in 1st. post in this thread, "No consensus yet". These wu is
> > re-issued to either a consensus is reached, or errors out. It means, the
> > validator didn't find a pair of results to be "nearly similar".
> >
> > > 877653
> >
> > This wu should have been re-surrected then the fix for re-issuing
> inconclusive
> > wu was added beginning of December. But since from August, it's possible
> got
> > permanently corrupted in one of the crashes...
> >
> > > 8909744,8909744 which all show 4 sucessful results and nothing
> > > credited and showing pending still.
> > >

======

Since 2 days (12.28.04) every result returns with "pending" on two different machines (host: 368059 and 360502); at least three results per WU was success.
What is wrong, what is going on?
How can I see that two results are "enough similar"???

See here for those pending results (over 40 the last two days):
http://setiweb.ssl.berkeley.edu/results.php?userid=7806448&offset=40
For example this: http://setiweb.ssl.berkeley.edu/workunit.php?wuid=6884054
or: http://setiweb.ssl.berkeley.edu/workunit.php?wuid=6837692
on this results-page (there are more...).

Before (12.28.04) all was running fine with Seti, no changes was taken, and now...???
thanks for reply

---

In progress 1,837,168 wu
so lets say 4 hr per wu, but we will use 1 per day.
Thats 1,837,168 wu return per day devide by three - 612
So the validater has to process at least 612 wu per day looking at three results each. It just takes time. What if the above wu are the third results for a wu, now thats makes 5511504 wu it's has to review per day. the last result was only received today.

---

Fred
BOINC Alpha, BOINC Beta, LHC Alpha, Einstein Alpha

> > Sorry was a misstype should be 890974 and 802411 both of which show 4
> results
> > all with no errors

Hmm, ah, see the problem, these should be re-sent, but don't expect this to happen before the new database is up and running. ;)

>
> ======
>
> Since 2 days (12.28.04) every result returns with "pending" on two different
> machines (host: 368059 and 360502); at least three results per WU was
> success.
> What is wrong, what is going on?
> How can I see that two results are "enough similar"???
>

Even the status-page says otherwise, either the transitioner or the validator or both is backlogged, so these wu haven't been through the validator yet. Or, the replicate-database is lagging behind, so even the wu already is validated this isn't showing up on the status-pages yet.

For a little more info, look at the Technical News.

> For a little more info, look at the Technical News.
>
=====

I didn't right unterstand (or missunterstood)the statements there; Thanks for reply ;-)
(now I got it!)

---

Lets hope a new database would fix the issue with the validator falling behind (something that has been happening for several days now). If it's consistently unable to validate as quickly as people send the results up...and people don't back off the processing of SETI units (a problem that could get worse as more people transition from SETI classic), this is only a problem that could get worse.

Following your example (and of course all our computers having different performing processors can complete units in dif amounts of time, but for simplification), lets say the validator processes 500 units per day, but 613 are being uploaded.

Now on day 2, the day would start with 113 (613 - 500 units) un-validated. It would validate another 500 units, except another 613 would be added. As a result there would be 113 unvalidated work units from the previous day + another 613 to be validated. This would bring the total to validate up to 726. But only 500 get validated again, leaving now 226 units left to be validated on day 3, etc... This is a problem that could grow ever worse, with more unvalidated units left on the dBase to yet be validated. This is without even adding an increasing number of SETI classic users migrating over.

Hopefully more validators will be added to the mix or something?

---

Well, as Matt said ...

He is working on new hardware to stand up and hopefully this will ease the burden.

---

there is something wrong with this wu: 7107925 just look at the CPU times and at the CPU's mine pentium M is by far the slowest of the three, but returned the wu a lot faster than the other two so I guess this one won't get credit when number four turns in hs results.

---

> there is something wrong with this wu: 7107925 just look at the CPU times and
> at the CPU's mine pentium M is by far the slowest of the three, but returned
> the wu a lot faster than the other two so I guess this one won't get credit
> when number four turns in hs results.
>

How fast a computer crunches seti-wu isn't only dependend on how fast the cpu is running, but also on other components like memory, mainboard and OS, and for a lesser extent on other processes running at the same time.

Taking a little look on this wu, it's one of the new wu initially sent out to 4 users, but is tried validated after 3 results. Since the results still shows validation-state "initial", they've not been tried validated yet but is waiting in the backlogged validator-queue. So atleast for the moment, nothing indicates there's anything wrong with this wu, and all current result seems consistent since they're showing "-9 too many signals".

> there is something wrong with this wu: 7107925 just look at the CPU times and
> at the CPU's mine pentium M is by far the slowest of the three, but returned
> the wu a lot faster than the other two so I guess this one won't get credit
> when number four turns in hs results.
>

The pentium-m, especially the new dothan with 2MB L2 cache, is a pretty fast cruncher.
Don't let the low GHz fool you.
It's doing a lot more work per clock cycle than a P4, for example.

Regards Hans

---

> How fast a computer crunches seti-wu isn't only dependend on how fast the cpu
> is running, but also on other components like memory, mainboard and OS, and
> for a lesser extent on other processes running at the same time.
>

BTW, it seems other processes running might have a minimal effect. After I started running BOINC, I also started running folding@home in parallel. I have f@h set to a slightly higher CPU utilization (so it runs) but set back to only use 50% of the CPU so seti can also run.

Before and after, the amount of time SETI work units takes to complete is as follows:

- Naturally it takes them longer to complete, however
- the time boinc says it was running for does not increase at all In this case, when it says 1 sec has gone by, more time has passed in real time. I can only imagine this is taking into account time it ran on the CPU, else it should be showing much longer running times for a SETI work unit in the interface with folding taking half the CPU. It doesn't.

Also the claimed credit is still rather consistent (doesn't show any breaks from the former trend) with or without folding running.

And for the other poster, clock isn't everything. The Athlon 64, I'd have to check this again, but for a given performance rating (aka 3000+ etc), the actual clock might be below some of the higher clocked Athlon XPs, but the Athlon 64 is a real processing power house. It's not all clock...

---