Double post ... delete please

Last night, latish, I swapped my CPU only rig from the opti client 0.19 to 0.20.

It looks like the new client is faster.

But, I an getting zero credit awarded as this example of the results shows.

I'm taking a look at the server right now and seeing what's up with the WUs that are being flagged as invalid. Hopefully I can get it sorted out soon.

I might suggest that the purge "script" be changed to remove valid tasks faster than then ones that are invalid or error. There are times where I am not sure that I am likely to have caught errors here because the purge can happen so fast...

The only thing that gets me a little riled up, is the lack of discussion BEFORE the credit adjustment.

If that is any relief to you, I have discussed with Travis about that reduction. I think it is pure concidence that the reduction is completely compensated by the new ATI version ;)

So what is the result? One can now point every cross project parity fanatic to the credit multiplier matching the equivalent value of SETI by saying MW gets high credits on ATIs just because they are so much faster! Furthermore nothing has changed to the worse here, quite to the contrary. The app is now delivering better results even faster than before.

Yeah, OK CP - I get your point. Only thing is: I'm CPU only ATM, and my CUDA card don't like them new Nvidia drivers!! ;^).

Looks like I'm gonna have to stick an ATI card into my old box if I want to benefit from your hard work!!

Cheers,
Chris

---

Seejay **Proud Member and Founder of BOINC Team Allprojectstats.com**

;-) ..with my current computer I do more science by reading
your posts than by crunching WU`s.. ;-D So please continue to
discuss what ever you may think of. lol

Excellent observation. And with my computers I can read faster than you and do even more science than if I were doing WUs... <Grin>

/Me Waves Hello....

---

The only thing that gets me a little riled up, is the lack of discussion BEFORE the credit adjustment. As Ice says, he's got personal goals based on this project, and it's US who actually do the crunching - therefore, why doesn't the admin here give us space to express OUR opinions before making important decisions?

Actually, if you were following the news posts, a month or so ago I did post that we would be lowering credits sometime in the near future. No one just got up and complained about it until I actually did the lowering.

If you look at previous news posts, on July 21st:

July 21, 2009
I'll be updating the server daemons tonight with better result verification and updated particle swarm and genetic searches. There will probably be some downtime while I get everything up and running correctly. We've also been examining some issues we're having with our fitness calculation, so we'll be releasing a new application and its code to deal with that as well. It also seems we're having some excessive credit issues again so things are going to have to be tweaked (have fun flaming our forums :P).

So it's not really fair to say I didn't warn you. You had over a month to gripe about it :)

---

The only thing that gets me a little riled up, is the lack of discussion BEFORE the credit adjustment. As Ice says, he's got personal goals based on this project, and it's US who actually do the crunching - therefore, why doesn't the admin here give us space to express OUR opinions before making important decisions?

Actually, if you were following the news posts, a month or so ago I did post that we would be lowering credits sometime in the near future. No one just got up and complained about it until I actually did the lowering.

If you look at previous news posts, on July 21st:

July 21, 2009
I'll be updating the server daemons tonight with better result verification and updated particle swarm and genetic searches. There will probably be some downtime while I get everything up and running correctly. We've also been examining some issues we're having with our fitness calculation, so we'll be releasing a new application and its code to deal with that as well. It also seems we're having some excessive credit issues again so things are going to have to be tweaked (have fun flaming our forums :P).

So it's not really fair to say I didn't warn you. You had over a month to gripe about it :)

Actually I did complain with no answer: http://milkyway.cs.rpi.edu/milkyway/forum_thread.php?id=1010&nowrap=true#28154. And you also did not say how much of a reduction.

---

Doesn't expecting the unexpected make the unexpected the expected?
If it makes sense, DON'T do it.

Last night, latish, I swapped my CPU only rig from the opti client 0.19 to 0.20.

It looks like the new client is faster.

But, I an getting zero credit awarded as this example of the results shows.

I'm taking a look at the server right now and seeing what's up with the WUs that are being flagged as invalid. Hopefully I can get it sorted out soon.

Travis, in my case the invalid WUs were caused by BOINC v6.10.4 and were accompanied by the message:

Milkyway@home 9/14/2009 1:35:24 PM Output file de_constrainted_82_2s_5_228058_1252953073_0_0 for task de_constrainted_82_2s_5_228058_1252953073_0 absent

Going back to v6.10.3 solved the problem here.

As far as the credit lowering, this project still gives higher credit than any other. I like credits as much as anyone, but agree that they were too high. Keep up the good work. Thanks for a nice project.

Travis Tweaks with a sledge hammer

---

As far as the credit lowering, this project still gives higher credit than any other. I like credits as much as anyone, but agree that they were too high. Keep up the good work. Thanks for a nice project.

I think this is the point. Our credits were VERY high, and now they're still quite good (maybe even still a bit too high).

If it makes you guys feel any better, Dave Anderson wanted our credit multiplier to be down to 2.7 (the same as SETI's), but I held my ground on 5.4 for our double precision work :P

Either way we're using what I think should be the correct multiplier for double precision work, any other tweaks from here on out should be due to changes in the FLOP count of the stock application, and not drastic changes to the multiplier.

---

To go back to the the same RAC you had before the sledgehammer was used, just install the new optimised v0.20 app from CP...OMG, my little 3850 should actually increase its RAC!

Either way MW is still the best CPU paying project around at the moment - maybe only until AQUA gets going again though.

If it makes you guys feel any better, Dave Anderson wanted our credit multiplier to be down to 2.7 (the same as SETI's), but I held my ground on 5.4 for our double precision work :P

As I've said repeatedly, changing the credit around removes all capability of accurate intra-project metrics between participants. If I started out a year ago and got an average of 50 cr/hr and put in 2000 hours, I'd have 100,000 credits. If someone starts now and averages 25 cr/hr and puts in 2000 hours in the next year, they show as having 50,000 credits. Technically both users put in the same amount of time, but the appearance is that the person with the 100,000 credits "did more work".

David Anderson is trying to adhere to something that is fundamentally flawed, and so are you by following him. You people from the project side need to step up and tell him that he's got no clothes on...

-Brian

If it makes you guys feel any better, Dave Anderson wanted our credit multiplier to be down to 2.7 (the same as SETI's), but I held my ground on 5.4 for our double precision work :P

As I've said repeatedly, changing the credit around removes all capability of accurate intra-project metrics between participants. If I started out a year ago and got an average of 50 cr/hr and put in 2000 hours, I'd have 100,000 credits. If someone starts now and averages 25 cr/hr and puts in 2000 hours in the next year, they show as having 50,000 credits. Technically both users put in the same amount of time, but the appearance is that the person with the 100,000 credits "did more work".

David Anderson is trying to adhere to something that is fundamentally flawed, and so are you by following him. You people from the project side need to step up and tell him that he's got no clothes on...

-Brian

I think everyone's aware that the credit system has some serious flaws and it needs to be addressed. What BOINC as a whole really needs is someone actively working on it.

It's definitely beyond the scope of what I can do, especially since I'm very busy trying to finish my PhD thesis this semester.

For any of you students out there (or at home researchers), something along the lines of "Fairly Rewarding Computational Effort in Volunteer Computing Grids" sounds like a good PhD thesis and/or set of publications :P

Honestly, there's no easy fix that will magically give all projects credit equality for work, otherwise something would have been done by now. It's really a pretty interesting research topic if anyone wanted to do something about it.

---

Travis Tweaks with a sledge hammer

Nothing wrong with a fine Sledgehammer... *grin*

PS. My ATI card just loves Gipsels new app! No real loss there - and even the CPU units still give good creds. ;-)

---

Lovely greetings, Cori

As I've said repeatedly, changing the credit around removes all capability of accurate intra-project metrics between participants. If I started out a year ago and got an average of 50 cr/hr and put in 2000 hours, I'd have 100,000 credits. If someone starts now and averages 25 cr/hr and puts in 2000 hours in the next year, they show as having 50,000 credits. Technically both users put in the same amount of time, but the appearance is that the person with the 100,000 credits "did more work".

You're trying to make credits out to be something like "time spend doing work" when it's really not that simple. Credits are awarded for the amount of work done (100 points = 86400e9 single precision floating-point operations AFAIK). Better hardware makes for better credit rates. Also, the amount of calculation done by the milkyway application has been decreasing over time as it gets more and more optimized. A user can definitely put in more time, but do less work. And in your example, due to the optimizations in the milkyway application, the older user most likely did do more work.

---

Just wanted to say hello. I just join MW. I've had a mild interest in the project for some time. The fact that an effort is being made to achieve credit parity is one of the reasons I decide to add the project to my list at this time. My contribution may be miniscule compared to others, but I thought that the effort should be rewarded.

---

Questions? Answers are in the BOINC Wiki.

Boinc V7.0.27
Win7 i5 3.33G 4GB, GTX470

If it makes you guys feel any better, Dave Anderson wanted our credit multiplier to be down to 2.7 (the same as SETI's), but I held my ground on 5.4 for our double precision work :P

Figures, any Project that gives out more Credit than SETI is on DA's Hit list to try and lower them to SETI Standards. But on the other hand they don't try to raise the Projects that give out less than SETI ... Go Figure

As I've said repeatedly, changing the credit around removes all capability of accurate intra-project metrics between participants. If I started out a year ago and got an average of 50 cr/hr and put in 2000 hours, I'd have 100,000 credits. If someone starts now and averages 25 cr/hr and puts in 2000 hours in the next year, they show as having 50,000 credits. Technically both users put in the same amount of time, but the appearance is that the person with the 100,000 credits "did more work".

You're trying to make credits out to be something like "time spend doing work" when it's really not that simple. Credits are awarded for the amount of work done (100 points = 86400e9 single precision floating-point operations AFAIK). Better hardware makes for better credit rates. Also, the amount of calculation done by the milkyway application has been decreasing over time as it gets more and more optimized. A user can definitely put in more time, but do less work. And in your example, due to the optimizations in the milkyway application, the older user most likely did do more work.

You're using the rationalization of what has transpired here with this single project, not "the big picture". I'd still challenge you to provide definitive proof of which person "did more work". Most people are going to look at the total amount in the credit bucket and make the determination. That's how the stat sites work...

Also, what about a period of time where the performance of the application remained constant? The new user has to do more than the old user who got the benefit of the higher rate...

The best thing for all of you that head up the various projects to do is to freeze the current credit system stats and start over with the "1 credit for 1 completed task" methodology, followed by official announcements by all of the projects as well as the BOINC staff that comparing projects is no longer supported by the BOINC platform and that the ONLY rankings that are supported are those within a single project. Each project is its' own entity, sharing only a common software framework. From that point on, if BOINCStats, BOINC Combined Statistics, All Project Stats, Knights Who Say Ni, or any of the other stat sites decide that they want to try to make an exchange system, IT IS ON THEM AND NO LONGER ON YOU, THE PROJECT, TO TRY TO KEEP UP WITH...

For the life of me I don't know why that doesn't sell with you project-side folks... It'd be one continual thorn in the side that you could get rid of... Instead though, you cling to the ways of the past, with a system that, while was a good idea in concept, had a horrible implementation...

If it makes you guys feel any better, Dave Anderson wanted our credit multiplier to be down to 2.7 (the same as SETI's), but I held my ground on 5.4 for our double precision work :P

Figures, any Project that gives out more Credit than SETI is on DA's Hit list to try and lower them to SETI Standards. But on the other hand they don't try to raise the Projects that give out less than SETI ... Go Figure

I'll add on to that.

It doesn't matter the amount of work done per task only the credits that it gives out.

---

Doesn't expecting the unexpected make the unexpected the expected?
If it makes sense, DON'T do it.

For the life of me I don't know why that doesn't sell with you project-side folks... It'd be one continual thorn in the side that you could get rid of... Instead though, you cling to the ways of the past, with a system that, while was a good idea in concept, had a horrible implementation...

I'm not clinging to anything. I'm just working with what we have. If anyone comes up with a better credit system we'd be more than happy to use it.

If we never lowered our credit from when the project first came out, each WU would be at about 20,000 credit. That's pretty ridiculous.

I honestly don't know what you want from us. People always complain about credit but never offer any suggestions. I personally don't have a solution.

There's no way to tell how many actual flops any workunit actually took, given the fact that there's a slew of optimized applications out there, running on all different kinds of hardware. Since credit is inherently tied to something un-calculateable, it's pretty hard to come up with any reliable way of determining it.

So what are we doing? We take our stock application, get a general idea of how many flops it will take from the code, (and now after the credit change) we're applying that to the supposedly standard credit multiplier, modified for double precision work.

What do you suggest we should do?

---

The best thing for all of you that head up the various projects to do is to freeze the current credit system stats and start over with the "1 credit for 1 completed task" methodology, followed by official announcements by all of the projects as well as the BOINC staff that comparing projects is no longer supported by the BOINC platform and that the ONLY rankings that are supported are those within a single project. Each project is its' own entity, sharing only a common software framework. From that point on, if BOINCStats, BOINC Combined Statistics, All Project Stats, Knights Who Say Ni, or any of the other stat sites decide that they want to try to make an exchange system, IT IS ON THEM AND NO LONGER ON YOU, THE PROJECT, TO TRY TO KEEP UP WITH...

But not all of our tasks take the same amount of work (compare 1 stream to 2 stream to 3 stream), does 1 credit for 1 completed task really make sense for that?

Also, what about projects with tasks that have nondeterministic runtimes?

At any rate, if this is what you want it's not too hard to apply some kind of multiplier to whatever credit we, or any other project, is giving out, such that it works out to 1 credit for 1 task. It wouldn't really matter then if we were giving out 1 credit per task or 700. This kind of solution is pretty independent to whatever credit we're giving out.

Also I think the BOINC server software already tracks how many total workunits a user has crunched. You could just take that information and ignore credit all together.

If anyone actually wanted to do something about this (instead of just complain about credit), it would only be a few extra lines of code to add that information to the stats export.

If you wanted to set up some kind of credit exchange/tracker site I'm sure many BOINC projects would be happy to upgrade their server code and get on board, and I'm sure that code could be added to the code trunk pretty easily, if it's not already being exported.

---