If Seti could come to a standard crunch time to the WU the classic setup would work. 1 credit per WU. I know it wont happen cause of the implementation of the new credit system throughout Boinc. But whatever, wouldnt be nice to have somthing that was simple as the setup that made Seti@home what it is today.

If Seti could come to a standard crunch time to the WU the classic setup would work. 1 credit per WU. I know it wont happen cause of the implementation of the new credit system throughout Boinc. But whatever, wouldnt be nice to have somthing that was simple as the setup that made Seti@home what it is today.

Sorry, this did not even work well in Classic. Even then, the different angle ranges had different crunch times, and people were motivated to write clients that would discard the results that were likely to take longer. So, this is a non-starter.

---

BOINC WIKI

With all the variance of the crunchtimes of workunits, not only at Seti but also at the other projects controlled by Boinc one would get the same credit whether his machine took 1/2 hour or 100 hours to crunch a single unit. And there just may be that much variance throughout all the projects run by Boinc, excluding the contaminated units.

---

With all the variance of the crunchtimes of workunits, not only at Seti but also at the other projects controlled by Boinc one would get the same credit whether his machine took 1/2 hour or 100 hours to crunch a single unit. And there just may be that much variance throughout all the projects run by Boinc, excluding the contaminated units.

Actually, the range is even wider than that. On the same computer, one project I run has results that take ~10 minutes, and another has results that take about ~2000 hours.

Even just in S@H results can take 4 hours or 40 hours with different angle ranges on the same computer. This is specifically excluting the noisy results (still valid science - we do need to know about them) that complete in under 5 minutes.

---

BOINC WIKI

Hi James,

I think the new Enhanced application is the closest we've come so far to emulating the Classic way of granting credit, with the additional capability of compensating for workunits of varying size. Whether you have a slow computer or a fast one, use an optimized cruncher or not, you will now receive the same credit as your peers. Not just in quorum, but across the board now.

This I feel puts the competitive aspect of SETI back to where it was on Classic... tuning your system hardware and software to crank out a greater number of workunits to gain a competitive edge.

There are bugs to be worked out for sure, and the amount of credit awarded via the Fpop mulitplier seems to be a bone of contention, but I think most everyone agrees that the Enhanced system itself is quite fair.

Dig

I'm not exactly sure if this has already been covered. I have tried to keep up with the discussions, but there is way to much out there.

I have always been a fan of assigning fixed credits based on WU. The SETI (or any) development team knows their code and they know their data. I would think that the development team could have some metric to assign credit to a given WU. Instead of the 1 credit per WU, there could be a credit value based on angle range (or angle range plus other factors.)

It seems that the multiplier method is a bit kludgy to me. It appears to be a metering sort of solution, which is appearing to be difficult across platforms. So, son't worry about it. There may be small differences, but it's better then the current situation. It would also help the optimized client situation. I admire Crunch3r's and others's efforts...I'm an ex-assembler programmer myself and believe in using hardware capabilities to their fullest. This should give them the freedom to concentrate on data accuracy, precision and efficiency without having to deal with credit issues.

I hope Crunch3r comes back soon. His and the many other experts work embodies all the best qualities of collabrative computing and is valuable to the scientific community. I think most of us would continue to contribute as long as the credit system is reasonably fair. Making it a back-end process can only help.

---

TheMule

It should be based on work done. (I am not familiar with the current formula.)

So, what would be the variables in a fair equation to award credit?

---

Fear will keep the local systems in line. Fear of this battle station. - Grand Moff Tarkin

It should be based on work done. (I am not familiar with the current formula.)

So, what would be the variables in a fair equation to award credit?

Depends on your definition of work ;)

kWh the computer uses is one way, but far too physicist, and would only be valued by those with AppleII or 386.

"Scientific value" would be right, but how to measure?

If the WUs behave in a very similar way, you can crunch one on a benchmark computer and predetermine the value, like it's done in CPDN and Folding. But this won't work with too diverse WUs, like Seti with extreme ARs, or LHC.

Flopcount would be fine, it's "the real stuff" that's to be done.

I'd prefer this all to be done by the client, as this is the comprehensive part of BOINC, but if it's fine and easy for the smaller projects to implement it in their app-code (like Seti did), it's fine with me as well.

But all this things need to be discussed on all projects, not just a single one.

---

Gruesse vom Saenger

For questions about Boinc look in the BOINC-Wiki

If the WUs behave in a very similar way, you can crunch one on a benchmark computer and predetermine the value, like it's done in CPDN and Folding. But this won't work with too diverse WUs, like Seti with extreme ARs, or LHC.

I believe Einstein@home plans to switch to a method like this when their “S5” project goes online, in a few weeks. There is variation in the WUs, but only to the extent that they come in two or three distinct sizes, so pre-evaluating them should be quite feasible.

---

The way I see it, we're all in this together. That is to say, whatever the method of determining and granting credit happens to be at the moment, it effects all of us equally over the course of time. Like most, I'm down about a third in RAC since switching to enhanced, but so is everybody. It would have been nice if enhanced were a new project the way some of the CPDN offshoots like BBC or Seasonal are so that credits stay straight for different applications, but in the end it doesn't really matter.

Sometimes credits get a little whacky in a project and whoever is chasing credits can take advantage of the normally short-lived phenomenon wherever it happens to be. Case in point, Einstein was giving arguably the lowest credit per computing hour of all BOINC projects but now, with Akosf's app coupled with a calibrating client, it probably gives the highest. That will change when S5 is released and they go to universal credit per type of WU, but I certainly hope they don't take a hit in terms of users over the change. Their credit system was designed to assume processing times of the stock app, so it's not really wholly compatible with a 400% improvement in terms of cross-project comparison.

Likewise, Cruch3r's 4.11 app took one of my machines from RAC 180 under the standard 4.18 to about 620. Now it's hanging around 400. Should I complain that it's not 610 or be grateful and remember that it used to be 180?

There has to be a way to determine the effort required to process a WU. Does the angle indicate the dificulty to process?

---

Fear will keep the local systems in line. Fear of this battle station. - Grand Moff Tarkin

There has to be a way to determine the effort required to process a WU. Does the angle indicate the dificulty to process?

Partly: it determines which tests are performed on the data, and is being used to calculate deadlines. I believe, though, that the actual amount of work required depends on how many potential signals are found, which is essentially unpredictable.

---

Hi Guys,

Although I'm a supporter of SETI Enhanced and the additional sensitivity it brings to the science, I think everyone involved in the testing of it (including Eric) will admit that there are flaws on the credit-side. I do think we are on the right track though in awarding credit based on work done rather than the old benchmarking system. It's a step in the right direction and the fairest BOINC credit calculation we've seen thus far.

I have to add that I am quite happy to see that this discussion has not erupted into yet another Berkeley-bashing thread. These guys have no motive whatsoever for being malicious or for getting 'greedy' with the credits. If they wanted to, they could assign any arbitrary large number they wanted to the multiplier and give everyone a windfall profit on credits. But as the developers of not only the SETI application but the BOINC platform as a whole, they have a responsibility to keeping things equitable cross-project as well as in-house.

Nice discussion! A lot of good ideas have been presented here.

Dig

Well, since this thread appears to be open minded, I will repeat what I have said before. The current system would have been a lot easier to swallow for the power users if they had simply reset everybody to zero when bringing it onboard. They could have captured the old scores as they did for classic and gave everybody a fresh start. That would have been a "true" level playing field. You have to admit that it was a major change in the way credit was figured and awarded, just as Boinc was a major change from classic. Not doing so was like getting your legs knocked out from under you for many of us.

---

There has to be a way to determine the effort required to process a WU. Does the angle indicate the dificulty to process?

Partly: it determines which tests are performed on the data, and is being used to calculate deadlines. I believe, though, that the actual amount of work required depends on how many potential signals are found, which is essentially unpredictable.

So potential signals times what should equal credit claimed?

---

Fear will keep the local systems in line. Fear of this battle station. - Grand Moff Tarkin

Well, since this thread appears to be open minded, I will repeat what I have said before. The current system would have been a lot easier to swallow for the power users if they had simply reset everybody to zero when bringing it onboard. They could have captured the old scores as they did for classic and gave everybody a fresh start. That would have been a "true" level playing field. You have to admit that it was a major change in the way credit was figured and awarded, just as Boinc was a major change from classic. Not doing so was like getting your legs knocked out from under you for many of us.

There allways was a true level playing field for everybody, and that was:

Benchmark x CPU-time / 1728000 = claimed credit

The benchmark had to be done in a fashion to match the one done by the original client on an average computer.

It was agreed upon, that for the really fast optimizers it would be OK for some time to adjust the benchmarks to claim the same as everyone else, but of course only do so with an optimized application running.

Imho if once a better solution would be found, this interim benchmark tweaking, that is highly cheating attracting, should be abandoned. The flop-count seemed to be the better solution, and once the factor to match the 1728000 was found, it was released.

---

Gruesse vom Saenger

For questions about Boinc look in the BOINC-Wiki

Well, since this thread appears to be open minded, I will repeat what I have said before. The current system would have been a lot easier to swallow for the power users if they had simply reset everybody to zero when bringing it onboard. They could have captured the old scores as they did for classic and gave everybody a fresh start. That would have been a "true" level playing field. You have to admit that it was a major change in the way credit was figured and awarded, just as Boinc was a major change from classic. Not doing so was like getting your legs knocked out from under you for many of us.

Hiya Sarge,

I follow what you are saying but I'm not sure I see why you would want that. As it is now, you are accumulating credit under the new system on top of what you have already earned to date under the old system, thus keeping your credit-position in relation to other crunchers here on SETI. Your RAC has dropped, but I assume that it has dropped in proportion to other crunchers in your group. If they zero'd everything out, would the end result not be the same only with smaller numbers? It seems to me that you would actually lose ground by forfeiting the credits you earned under the old system.

Perhaps I am just missing something here and of course I give you the benefit of the doubt, so if there is some key fact I am missing here please let me know. I am admittedly no mathematical genius :)

Dig

Well, since this thread appears to be open minded, I will repeat what I have said before. The current system would have been a lot easier to swallow for the power users if they had simply reset everybody to zero when bringing it onboard. They could have captured the old scores as they did for classic and gave everybody a fresh start. That would have been a "true" level playing field. You have to admit that it was a major change in the way credit was figured and awarded, just as Boinc was a major change from classic. Not doing so was like getting your legs knocked out from under you for many of us.

There allways was a true level playing field for everybody, and that was:

Benchmark x CPU-time / 1728000 = claimed credit

The benchmark had to be done in a fashion to match the one done by the original client on an average computer.

It was agreed upon, that for the really fast optimizers it would be OK for some time to adjust the benchmarks to claim the same as everyone else, but of course only do so with an optimized application running.

Imho if once a better solution would be found, this interim benchmark tweaking, that is highly cheating attracting, should be abandoned. The flop-count seemed to be the better solution, and once the factor to match the 1728000 was found, it was released.

That is your opinion and I respect it. But I was talking for the power users. At my peak I was up to 10K credits a day. In one fell swoop I have dropped to below 3K. This is a drastic drop for those of us who were competing for the top spot. Competition was the driving train for the program and even Dr. Werthimer stated in an interview that the competitive side amazed even him for the amount of computing power it brought to the project. Now the project may lose many of the power users who enjoyed it for the competition. Kind of defeats what Dr. Werthimer was hoping for.

---

That is your opinion and I respect it. But I was talking for the power users. At my peak I was up to 10K credits a day. In one fell swoop I have dropped to below 3K. This is a drastic drop for those of us who were competing for the top spot. Competition was the driving train for the program and even Dr. Werthimer stated in an interview that the competitive side amazed even him for the amount of computing power it brought to the project. Now the project may lose many of the power users who enjoyed it for the competition. Kind of defeats what Dr. Werthimer was hoping for.

But Credits is a measurement for valid science delivered, so you have delivered more valid science then others in that time, I assume through an optimized app and client.

Now the playground is levelled again, and those with an optimized client don't get that much bonus any more, as they don't deliver that much more then average for that certain puter.

There are still some issues with hanging WUs and C++ errors, but imho it's fine if the project will use somehow optimized clienz´ts for all, and thus make the extra bonusses for the "pioneers" obsolete. Don't discard them, they were correct at the time granted, but don't go on with this inflation of credits.

And the "benchmark" for a good credit system is still: will it bring back the level playingfield once set by the original equation?

Well, since this thread appears to be open minded, I will repeat what I have said before. The current system would have been a lot easier to swallow for the power users if they had simply reset everybody to zero when bringing it onboard. They could have captured the old scores as they did for classic and gave everybody a fresh start. That would have been a "true" level playing field. You have to admit that it was a major change in the way credit was figured and awarded, just as Boinc was a major change from classic. Not doing so was like getting your legs knocked out from under you for many of us.

I understand completely what your saying and tend to agree with you but we need to look at the realities of the past performances of participants on this project as well. If what you propose was done, we'd all still be here yapping about the same exact thing as we are now.

Everyone and their uncle would be complaining about the new system and claiming that they have a better way to do it. Power users would be complaining about their high credits and RAC's being wiped (even though they would be preserved as in the manner Classic totals were based on your proposal) as well as having to start over from scratch. Non-power users would still be criticizing the power users. Power users wouls still be criticizing the non-power users. Cries from both sides would continue about how unfair things are just as they are now.

In short, nothing that's done would satisfy everyone and end this.

---

To truly explore, one must keep an open mind...