After my first day on BIONIC SETI, I have more questions than anyone should have who has been running SETI since 1999.

Why is my credit reduced between the time the WU is returned and credit is finally "granted" (like it's some sort of gift, rather than something I earned by spending my cycles).

Outside of the mysterious way that the orignal "claimed" number is calculated (why not just "1" for a completed WU?), then for some unknown reason it is adjusted to a lower number. What is going on?

I can't get any straight answers on the Q&A forum, only insults, so I thought I would ask on this forum.

---

Hello again, and welcome Angus!

All WUs have different size, some stop after only 200sec, because they are too noisy. So a WU is not a WU, the real measurement is flops.
But to count them is a bit overdone. So they are calculated with the CPU-time and the benchmark of your puter, and some factor to get them more usable, and so we get this:

claimed credit = ([whetstone]+[dhrystone]) * wu_cpu_time_in_sec / 1728000

Next thing is the validation. There is no need to give rewards to junk. That way cheaters and people with cranky hardware/OS setup, who do not crunch properly will not get any credit. That's fine with me.
Validation has do be made in comparsion to the other results for the same WU. How much redundancy is needed is for the science team to calculate/adjust. Atm it's 3, so 4 are sent out, as some will not come back either in time or just invalid.

Then the credit will be granted. As the machines will all claim some different credit, the highes and lowest claimed are discarded, and the remaining 1or 2 are averaged.

The benchmarks part of Boinc is still improvable, but it is quite OK with the newer 4.2x clients. In some weeks no more of the old 4.19 will be out there, and the claimed credit will be more in line as it is now.

And, as I said over @Help Desk:

The (very good) manual for Boinc is the BOINC Powered Projects Documentation by Paul D. Buck and the FAQ therein.

For information regarding Credits look here!

---

Gruesse vom Saenger

For questions about Boinc look in the BOINC-Wiki

Hi Angus, and welcome to BOINC!!

Yes, there are big differences in calculating credit now, as you can clearly see... the amount of credit your computer claims is representative of the number of floating point operations (FLOPs) that it took for your computer to complete the unit. This amount is fount by multiplying the time that it took you by the numbers you got during your most recent benchmarks (they are done every time you load up BOINC or every 5 days, which ever is more often)

WHY??? you ask. Very simple. This is indicative of the amount of work that was done by your computer. As you know every so often there are very short (less than 5 minute) WUs because of "noise"... in classic if you were "lucky" enough to get a bunch of these, then you could get a huge number of WUs completed in a very short time.

However, with this method, getting "lucky" like that doesn't occur; I have WUs where I got 0.38 credits... cause it took me something like 2 min before they were declared as static filled.

Similarly, in Classic if you were "unlucky" and got a really long WU, you'd only be able to complete one while normally you could complete 2 or 3. With BOINC, you actually GET 3 times the credit.

The reason for credit being granted after a delay rather than immediately is to ensure that the results that are being reported are consistent. Each WU is sent out 4 times and once three have been properly completed, they are compared to each other by the validator. If they agree, credit is granted, and if they do NOT agree, additional copies of the WU are sent out (to a max of 8 or so copies) until 3 are returned that agree. This is to ensure that the results that are returned are scientifically sound.

Over clocking your computer, for example, may speed up the rate at which you return units, but can also cause errors in computing, so if you OVER-overclock, your results will not agree with everyone else's, and you will not be granted credits.

Depending on the project, the exact method of determining the credits varies. If I recall correctly, in SETI, once the units have been validated, the system throws out the top and bottom values and grants the arithmetic average of any remaining results.

Therefore, sometimes you will be adjusted up, and sometimes you will be adjusted down, and sometimes you will get what you say you deserve.

I hope that this is helpful. My other suggestion for answers to all your BOINCing questions is to go to the BOINC owner's Manual, which has been compiled by Paul Buck (and will probably also respond to this question, knowing him)... There's also a FAQ on his site, but I can't find the URL right now.

Again, welcome to BOINC, and although there are growing pains when changing systems, I hope you see that there are many improvements over SETI Classic. =-)

---

> After my first day on BONIC SETI, I have more questions than anyone should
> have who has been running SETI since 1999.

Actually, probably not.

BOINC is very different from classic. It does a lot of things that classic could only do with a lot of help, and it doesn't allow some things at all that classic did easily.

It also has a number of features to try to fairly grant credit -- and to grant similar credit across more than one project.

So, yeah, you're going to have questions about differences, and unlike a SETI newbie, you're going to expect things to be a certain way when they aren't.

Welcome to the group....

---

> Hello again, and welcome Angus!
>
> All WUs have different size, some stop after only 200sec, because they are too
> noisy. So a WU is not a WU, the real measurement is flops.
> But to count them is a bit overdone. So they are calculated with the CPU-time
> and the benchmark of your puter, and some factor to get them more usable, and
> so we get this:
>
> claimed credit = ([whetstone]+[dhrystone]) * wu_cpu_time_in_sec /
> 1728000
>
> Next thing is the validation. There is no need to give rewards to junk. That
> way cheaters and people with cranky hardware/OS setup, who do not crunch
> properly will not get any credit. That's fine with me.
> Validation has do be made in comparsion to the other results for the same WU.
> How much redundancy is needed is for the science team to calculate/adjust. Atm
> it's 3, so 4 are sent out, as some will not come back either in time or just
> invalid.
>
> Then the credit will be granted. As the machines will all claim some different
> credit, the highes and lowest claimed are discarded, and the remaining 1or 2
> are averaged.
>
> The benchmarks part of Boinc is still improvable, but it is quite OK with the
> newer 4.2x clients. In some weeks no more of the old 4.19 will be out there,
> and the claimed credit will be more in line as it is now.

What I don't understand at all is why you would go to all the trouble to calculate some arbitrary credit number for each PC, for each WU, then throw all that away and assign the same credit to everyone who turns in a good result for that WU.

Why not just give them a "1" for returning a good one, and a "0" for an unusable one. It worked in Classic (except that you at SETI could never figure out how to check for duplicates and bogus returns - DOH). The whole process would run a lot faster and smoother, and the stats questions would certainly trickle off to almost nothing.

This is complication for the sake of obfuscation, I think. K.I.S.S. is called for here.

---

> > After my first day on BONIC SETI, I have more questions than anyone
> should
> > have who has been running SETI since 1999.
>
> Actually, probably not.
>
> BOINC is very different from classic. It does a lot of things that classic
> could only do with a lot of help, and it doesn't allow some things at all that
> classic did easily.
>
> It also has a number of features to try to fairly grant credit -- and to grant
> similar credit across more than one project.

What does this have to do with anything? I'm only talking about SETI here. There is no reason why anyone would want to compare SETI results with any other project results, in BIONIC or any other platform.

---

> After my first day on BIONIC SETI, I have more questions than anyone should
> have who has been running SETI since 1999.

Learning something new can take time, regardless of experience ;) I had many when I started on boinc as well - and I actually find this much more fun - there's so many more things to learn about boinc than there was with classic. To me, that's more fun ;)

> Why is my credit reduced between the time the WU is returned and credit is
> finally "granted" (like it's some sort of gift, rather than something I
> earned by spending my cycles).

I wouldn't say it get's reduced ;) You could regard copplestones (the credit system used on boinc) as a currency and the results as a commodity you want to sell to the server. When 3 salesmen want to sell equal results to the server, they all put a price on their results. Naturally the server want to pay with equal amount of the currency for equal results. So it evaluates the value of the results and hold that up against the price claimed by the salesmen. Just and example ;)

> Outside of the mysterious way that the orignal "claimed" number is calculated
> (why not just "1" for a completed WU?), then for some unknown reason it is
> adjusted to a lower number. What is going on?

Well, others have already explained the "why not 1 for completed wu" ;) I might add that it is a much more fair system. Another example:

You work at a factory, 12 hours a day. A collegue of yours only work 2 hours a day. The thing is, although you work much harder than your collegue, he gets the same salury is you do. You get 120 currency for a days work, and so does he. The type of work you do is excactly the same. It's not like he get's the same amount of work done in two hours as you do in twelve, actually you do the same amount of work in two hours as he does - only in a day you end up doing 6 times the work he does. Now, wouldn't it be more fair if he then only got 20 currency for a days work ? ;)

Hope this helps a bit :)

---

>
> What I don't understand at all is why you would go to all the trouble to
> calculate some arbitrary credit number for each PC, for each WU, then throw
> all that away and assign the same credit to everyone who turns in a good
> result for that WU.

The current benchmark isn't perfect, and some users will try to cheat with artificially increasing their benchmark-score and therefore claiming more credit.

Results must in any case be validated to be scientifically usable, so either issuing lowest claimed or average after highest/lowest claimed removed guards against users trying to claim too much credit.

>
> Why not just give them a "1" for returning a good one, and a "0" for an
> unusable one. It worked in Classic (except that you at SETI could never
> figure out how to check for duplicates and bogus returns - DOH). The whole
> process would run a lot faster and smoother, and the stats questions would
> certainly trickle off to almost nothing.

BOINC is not just SETI@home, you have LHC@home with a short wu taking around 30 minutes to a CPDN-wu taking 1 month. Also, all projects has already multiple wu-types or is expected to have multiple wu-types in some months time, so even if you're running only one project a wu-count wouldn't really be a good measurement.

Just for further reading:

The Seti/Boinc VS Seti@home Classic thread

or even this one over @Einstein ;)

there are several threads on the boards related to this very thing

i don't know if this will help or not but here goes:

your computer is going to "claim credit" based on the amount of time that it took to crunch "X" work unit. your getting that amount of credit will be determined by how much time it takes everyone else (that it was sent to) to crunch it. it may be more than what you claim or less.
there are fast and slow pc's out there. if a WU is sent to say 3 fast pc's and yours is slower then you will probably have a higher claimed credit than they will.
i would try to explain the benchmarking process as it is part of the process, but i think someone else could probably give a better and shorter explanation than i could :-)

btw, welcome!


.

---

No matter where you go, there you are...

> There is no reason why anyone would want to compare SETI results with any
> other project results, in BIONIC or any other platform.
>
>

For the same reason we do anything else, because we can. And so that we can remain competitive while doing multiple projects (a good thing, server outages don't stop a machine from getting WUs). It's just a different approach, and for better or worse, it's the next step.

---

Still looking for something profound or inspirational to place here.

> What does this have to do with anything? I'm only talking about SETI here.
> There is no reason why anyone would want to compare SETI results with any
> other project results, in BIONIC or any other platform.

Actually, there is ;) You see, boinc makes it possible to run more than just one project on the same computer. And when you do that, it can be a problem to compare yourself with others in the statistics. So there are sites that create cross-project statistics where you can compare your progress with that of others regardless of which and how many other projects you participate in.

But this means there should be a common ground for all those projects. Using wu count is totally useless for this, as avarage computation time for a wu can vary extremely from project to project. On my computer a seti wu takes about 4 hours to complete, while a cpdn wu takes about 25 days to complete. And that is only taking pure cpu time into consideration.

I know there are many people that only want to run seti, and don't see the point - but there are just as many that want to run other projects as well, as there are many very interesting projects outthere. There are also those who run other projects that want to run seti as well. And that's basically why a cross-project credit system is needed.

---

> What does this have to do with anything? I'm only talking about SETI here.
> There is no reason why anyone would want to compare SETI results with any
> other project results, in BIONIC or any other platform.

It's BOINC, not BIONIC.

Okay, let's limit just to SETI, and talk about the future.

Currently we get work from one telescope, in Puerto Rico.

At some point, other telescopes will start feeding work, and the work may be different.

While the application processing that work may still be "SETI" it will likely be a totall different kind of work -- it may process quickly, or take twice as long.

There are higher bandwidth studies, like Astropulse, that will be different work.

We'd like to be able to compare results for vastly different kinds of science even within the same project.

... and those quick, noisy work units that inflated scores in classic aren't worth much in BOINC/SETI.

So this feature will be important even if you only do SETI.

---

Overall it's a much more fair system than Classic i think, although as Ingleside stated the current benchmark isn't perfect. Faster machines tend to get penalized slightly and receive lower credit per work unit on average than a slower machine. But these are things that hopefully will get ironed out at some point in the future.

BOINC really is a lot of fun though, and once you give it a chance i think you'll like it quite a bit. :)

Dig

---

Angus, so far you've been given enough data and advice and opinions to answer the question "Can some actually EXPLAIN how credit is granted?". If you've read the pages on credit from Paul Bucks website, then you know as much as pretty much anyone. all this explains "How" and sometimes "Why" the Credit is the way it is.

Noone that has posted here WORKS at Berkeley. They (like I) can not change the program. We CAN explain "How" and sometimes "Why", BUT we can't change it. So If you're just not happy with it, then you need to address "Berkeley".

I wish you well

Welcome to BOINC (Berkeley Open Infrastructure for Network Computing)

hope this helps

tony

---

> I hope that this is helpful. My other suggestion for answers to all your
> BOINCing questions is to go to the <a> href="http://boinc-doc.net/index.php">BOINC owner's Manual[/url], which has been
> compiled by Paul Buck (and will probably also respond to this question,
> knowing him)... There's also a FAQ on his site, but I can't find the URL right
> now.

Heck, all I can do is add my signature with the link to the site.

You guys did a wonderful job explaining, so I cannot think of much to add ...

Angus,

I do want to say that this system was not designed on a whim, and there have been changes made to improve the system. For those of us that like to count work units completed you can do that. There are logging programs out there that will allow you to capture that data if you wish.

I agree, that is is sometimes painful to process a work unit and then to not be able to gain credit for the effort. WIth LHC@Home we had a long, long, fight with a bug that would zero out the CPU time. But the results, the real point of what we are doing, were still scientifically valid.

I have been doing BOINC for over a year and we have had these debates and discussions, but, so far, no one; and I mean no one (and that is including me) has come up with a system that is "fairer" and "better" than what we have now.

Is the system fool-proof? no
Is the system perfect? no
Is it better? I think yes it is. Because it is harder to "game" than most other systems.

When you look at my numbers you know without a single doubt, that I have done at least that much work. I can't report results that I did not get issued so I can't clone answers. We can't collude with you and I pooling our "answers" to inflate the work... The work returned has to agree with other answers or it does not count so I can't short change the calculations ...

But most importantly, the results are now subjected to more scientific rigor as a design goal of the system itself.

---

One more thing to add here...

Some of these wu's might give you a "gift", i.e.: you claim 20 and get 50

Some of them will leave you feeling like you've been robbed: claim 50 get 10



.

---

No matter where you go, there you are...

Well, there's been a lot of smokescreen put up about BIONIC and how it's better, because of all those other projects can be compared with SETI.

For someone from Classic SETI, who could not give a fig for those other projects, this just doesn't work. If the WUs change because SETI wants to analyze somehow differently, then everyone gets new WUs and the 1 credit = 1 WU still works. We all made it through the code changes in Classic when the crunch time was lengthened, because everyone was in the same boat. Put some time and money into fixing the hacker holes, and the project could continue like it was.

Trying to make SETI fit into someone else's nightmare framework is a disaster, and the concessions that were made are obvious. SETI/Berkeley would have been better off to put their time and money into redesigning their own project, instead of losing their identity in the BIONIC morass.

I guess I thought I was talking to Berkeley on these forums, since they are on a Berkely server, and linked from the SETI site.

Instead it seems like it's only a bunch of cheerleaders for BIONIC that have no say in how things are done, and can only see things one way. And you all wonder why less than 2% of the Classic SETI folks have made the switch...

I'm outta here.

---

> Instead it seems like it's only a bunch of cheerleaders for BIONIC that have
> no say in how things are done, and can only see things one way.
>
> I'm outta here.
>
Well, Like always... you're welcome to do what you wish. stay or leave it's up to you. Many fine folks are here to help others utilize Boinc to the betterment of Science. The cream of the crop have taken time to try to help you.

you're welcome for the assistance

have a nice nite.

Angus:

Nobody can make everybody happy. Heck, if we do find proof of the existence of ETI I can guarantee that alot of people around the world would be angry as h*ll - now the universe isn't what they've always believed it was, and that will scare them with anger as the response. Meanwhile all of us SETI@home participants will be extremely happy.

BTW: be careful as to which remarks you choose to express when you mention insults being made to your questions...

---