How many people are crunching Rosetta?
Message boards : Number crunching : How many people are crunching Rosetta?
Previous · 1 · 2 · 3 · 4
| Author | Message |
|---|---|
 rochester new york Send message Joined: 2 Jul 06 Posts: 2842 Credit: 2,020,043 RAC: 0 |
I have founded a graphic since 2007: http://boincstats.com/stats/project_graph.php?pr=rosetta |
Murasaki Send message Joined: 20 Apr 06 Posts: 303 Credit: 511,418 RAC: 0 |
I have founded a graphic since 2007: Comparing average credit from a few years ago to credit today may not tell you much. As "credit" is an arbitrary figure set by the project you would have to compare two time periods where the credit rules were the same. If, for example, the project team changed the rules tomorrow so that you earned double credits for the same amount of work, what would comparing the RAC for August 2009 against August 2008 tell you? The graph would show a massive climb in awarded credit, but could be hiding a decline in the number of active users. Likewise, a change that reduced the amount of credit awarded could hide an increase in the amount of work done. |
|
Sid Celery Send message Joined: 11 Feb 08 Posts: 2130 Credit: 41,424,155 RAC: 14,205 
|
I have founded a graphic since 2007: I'm not interested in credits but obviously some people are. I don't think that graph tells us anything much at all either. As an exercise: What I did see, though, was Project Credit Comparison where, if I'm reading it right, the second table shows hosts who run more than one project and the comparison of credits received for each (whether by WU or by hour of processing, I have no idea). Pretty consistently Rosetta 'pays' less than the majority of other projects (but more than one or two of the other bigger projects). I know there are people who do chase the higher paying projects and I have noticed that few months ago I was consistently getting granted more than claimed and now it's predominantly the other way round. I don't appreciate why it was high before, nor why it's low now, but I'd struggle to accept that the method being used was consistent. I'm about 15% down on RAC from where I was 2 months ago.  
|
|
Emigdio Lopez Laburu Send message Joined: 25 Feb 06 Posts: 61 Credit: 40,240,061 RAC: 0 |
Excuse me but I,m confused... I always though that credits (in Rosetta and in all the rest of the BOINC projects) were succesful works accepted by the team project to work with. I mean... credits are usefull information processed by the user and is valid for the project. If you have few credits means that your contribution for the project is low. Am I wrong?. Still not interested in credits, Sid? My "analysis" (sic) about these graphics is: - The power of the CPU,s (in general) has been dobled (or more) in the last 2 years - The power of process in Rosetta is more or less stable in these last two years This could mean that: - Rosetta users do not change their computers and the users are more or less the same :) or - R@H is not able to maintain volunteers. The new volunteers incorporated to the project has more CPU power but they spend less time with the project. In both cases the results is what I said before: The number of TFLOPS working for Rosetta (ideally) should be much higher than the actual 80-90 TFLOPS. The question is... why Rosetta is not able to increase the CPU power for the project?. What should be able to do Rosetta with 500TFLOPS (in theory, R@H should perform the actual work for 1 week in 2 days)? Kind Regards. 
|
|
Mod.Sense Volunteer moderator Send message Joined: 22 Aug 06 Posts: 4018 Credit: 0 RAC: 0 |
Am I correct to presume the chart is of average credit per user... per day? It could simply mean there are more projects available now and so the average user now has a lower resource share to Rosetta then they used to. If inactive users have not been omitted from the number, it could also mean many people have created new accounts for testing and other purposes over time and are no longer crunching on them so you are now averaging in a lot more zeros then you used to. If Rosetta had 500TFLOPs consistently available, it would basically mean that the researchers in BakerLab would all be able to more thoroughly explore their ideas at the same time, rather then waiting in a queue for the current projects to run through the system. It would also mean that a much larger sampling of each protein could be done. For example, what if I have an idea about improving Rosetta, we run it on 100,000 models for each of a dozen proteins and it only yields a slightly better model for one of them? Was it a good idea worth keeping or not? But what if this idea happens to only work well when a million models are available to draw from. And with that larger sample space it consistently produces measurably better results? There are always trade-offs. And when you know there are 100TFLOPs available to run the work, you tend to start thinking within that box. And so when you are designing new code, you tend to steer things into that reality, even if you actually sat down and thought it through, you would perfer to take the long road. So, having 500TFLOPs running instead would actually change the mindset of the researchers slightly. Perhaps create an environment where they try the long road more often. The whole problem to some extent is like trying to make a light blub. Edison said the experiment wasn't a failure, because we now one material that does not work and this narrows our focus going forward. With more computing power, you can try more ideas, faster. Rosetta Moderator: Mod.Sense |
|
ejuel Send message Joined: 8 Feb 07 Posts: 78 Credit: 4,447,069 RAC: 0 |
I'm sorry, I didn't read through this thread earlier, I wish I had... David...where is this report? Please provide the direct URL. It has been 2+ full weeks since your promise of a report. -Eric 
|
|
TomaszPawel Send message Joined: 28 Apr 07 Posts: 54 Credit: 2,791,145 RAC: 0 |
The question is... why Rosetta is not able to increase the CPU power for the project? I am fan of R@H, I was crunching this project on all my host's. But, I also like to have good RAC.... And when I see for 10,182.08s of work 54.44c .... I am SAD AND P@OFF... When I run MW@H for 10,182.08s of work I have 225c.... NO COMENT You have your answer! THANK YOU ROSETTA. WWW of Polish National Team - Join! Crunch! Win! |
|
rochester new york Send message Joined: 2 Jul 06 Posts: 2842 Credit: 2,020,043 RAC: 0 |
The question is... why Rosetta is not able to increase the CPU power for the project? |
mikey Send message Joined: 5 Jan 06 Posts: 1895 Credit: 9,208,737 RAC: 2,882 
|
Am I correct to presume the chart is of average credit per user... per day? It could simply mean there are more projects available now and so the average user now has a lower resource share to Rosetta then they used to. If inactive users have not been omitted from the number, it could also mean many people have created new accounts for testing and other purposes over time and are no longer crunching on them so you are now averaging in a lot more zeros then you used to. If you want LOTS AND LOTS of users here, give two or even 10 times the amount of credits as you do now per work unit!! There are TONS of credit whores out there that will flock here with their machines just to pad their numbers. Now that also means that a high percentage of those could be looking for a way to scam the system and get high credits for little or no work being actually done, so a high degree of checking would have to be instituted. Also if it takes 2 weeks for those credits to come thru, all those folks will fill the boards with complaints!! In short getting more users is not as simple as doubling or even tripling the amount of credits given out per work unit. YOUR workload Mod.Sense would also double or triple, as would the Servers, etc, etc, etc! 
|
|
Sid Celery Send message Joined: 11 Feb 08 Posts: 2130 Credit: 41,424,155 RAC: 14,205
|
Excuse me but I,m confused... I always though that credits (in Rosetta and in all the rest of the BOINC projects) were successful works accepted by the team project to work with. I mean... credits are useful information processed by the user and is valid for the project. If you have few credits means that your contribution for the project is low. Am I wrong? You certainly ought to be right, but I'd say it's not true. As a poor example I noticed something strange with my own credits this week. Previously I was getting about 50 creditsWU, with the granted credit about 30% less than claimed (actually 32.6% lower). In the last day or so, claimed credits almost halved (actually 43.7% down) but granted credits 37% higher than claimed. I don't know why. I did some sums and it turns out that granted credit gives me 1 credit per 288.65 seconds of CPU time on average, while last week it only averaged 273.71 seconds, so for the same contribution I get 5.17% less credit. I know this is an unrepresentative sample, but I do know that 6 months ago I was getting about 1600-1700 credits a day and now it's about 1250. 20-25% down. So no, credits have nothing to do with anything. They're just arbitrary numbers that don't even mean anything within a project and certainly have no meaning from one project to the next. Still not interested in credits, Sid? Personally, no. Not the slightest interest at all (as much as it might not seem that way!) If I was interested I'd be campaigning for Seti to get 0.01 credits per million WUs processed (and thinking that was over-generous too). A more worthless project I cannot imagine. What does interest me is that other people are interested in credits and are prepared to swap projects to chase those that are considered to over-pay. In that sense it's highly important, which is why I pointed at that project comparison graph showing Rosetta is less generous than others. In both cases the results is what I said before: The number of TFLOPS working for Rosetta (ideally) should be much higher than the actual 80-90 TFLOPS. Earlier you mentioned 3-400TFlops and now 500. Why not 800? This is meaningless. Credits should not reduce for the same amount of work being done. If my poor example shows that's the case then something should be done about it. If I wasn't getting 20-25% less credit than a few months ago, and this was reflected in other people too, we wouldn't have 80TFlops and have to listen to people grizzling about it.
|
|
ejuel Send message Joined: 8 Feb 07 Posts: 78 Credit: 4,447,069 RAC: 0 |
I'm not sure why people care about the credits at all...I mean, it's not like I can cash them in for an Amazon gift card or anything. Sure, it's nice to be in the top 5% or whatever of your project for bragging rights...but to jump from project to project to get some kind of "credit" gain just doesn't make any sense to me...there is absolutely no value. If people want to "belong" to a particular project just because it gives 3% more "credit" than another project for your identical client computer, then how does that benefit the person? Sounds pretty stupid if you ask me. Sheez...I mean I could invent a Credit Project that simply runs your CPU at 100% and counts clock cycles and gives you 5x the "credit" that any other BOINC project gives. Big whoop. So you get 6,000,000,000,000 points a day with my project...and???? People certainly have every right to belong to any BOINC project they want...but you would think there would be a little pride out there that you belong to a project because you believe it's a worthy cause for humanity. -Eric
|
|
mikey Send message Joined: 5 Jan 06 Posts: 1895 Credit: 9,208,737 RAC: 2,882
|
I'm not sure why people care about the credits at all...I mean, it's not like I can cash them in for an Amazon gift card or anything. Competition is not just a human thing, we do it to compare ourselves to others. We try to find out where we are in the whole scheme of things that way. You are starting to get into the mental health aspects of humans here and this could lead to a whole huge discussion about the human brain and its inner workings. Suffice it to say that my computers are better than your computers, so by extension I am better than you, is the thinking. Is this right, not in my opinion, but it is just my opinion!
|
|
Emigdio Lopez Laburu Send message Joined: 25 Feb 06 Posts: 61 Credit: 40,240,061 RAC: 0 |
So... if credits are not important for some of you... what alternative you have to measure the computational power working for R@H (or any other BOINC proyect)?. Bellow you have the actual statistics where you can see that Rosetta has more users than any other project (except SETI that not appears in this figures) but it is the 6th in recent credits. Project Total Credit Recent Credit Users Teams Hosts (Computers) Countries World Community Grid 15,317,451,841 28,401,559 247,75 15,58 758,02 218 Einstein@Home 13,905,674,053 17,684,637 235,78 8,7 968,48 214 ClimatePrediction.Net 8,587,799,501 16,897,744 199,13 6,61 385,13 213 Rosetta@Home 7,618,206,886 7,787,027 256,87 8,16 767,41 222 MilkyWay@home 5,466,152,242 45,117,900 29,23 1,22 65,29 158 GPUGRID 3,117,862,429 15,002,319 5,31 454 8,59 95 QMC@Home 1,958,628,043 2,071,955 37,16 1,8 87,82 169 ABC@home 1,916,734,486 2,821,914 22,56 1,14 62,16 151 BBC Climate Change Experiment 1,431,088,567 0 120,48 1,2 136,61 103 PrimeGrid 1,303,535,887 4,794,519 25,87 1,43 74,5 159
|
|
ejuel Send message Joined: 8 Feb 07 Posts: 78 Credit: 4,447,069 RAC: 0 |
Competition is not just a human thing, we do it to compare ourselves to others. We try to find out where we are in the whole scheme of things that way. You are starting to get into the mental health aspects of humans here and this could lead to a whole huge discussion about the human brain and its inner workings. Suffice it to say that my computers are better than your computers, so by extension I am better than you, is the thinking. Is this right, not in my opinion, but it is just my opinion! Oh I get it and understand it all...that's not my point. My point is that jumping from 1 project to another so you can get an extra X points is just dumb...in my opinion. The computer doesn't get any faster or better...you just get more points. Big deal. My computer can still run Ms. Pacman and I can get 200,000 points in a game...or I can play Asteroids and get 300,000 points in a game. My computer didn't get any better/faster...all I did was, in reality, alter/skew/cheat the system which somehow makes ME BELIEVE it is better.
|
|
ejuel Send message Joined: 8 Feb 07 Posts: 78 Credit: 4,447,069 RAC: 0 |
So... if credits are not important for some of you... what alternative you have to measure the computational power working for R@H (or any other BOINC proyect)?. The only importance of credit for me is: 1)I can be sure my machines are truly giving RAH results...if my RAC or credit is 0, I know something is wrong. :) 2)Within this project, sure, it is nice to have bragging rights...but really, to me, the bragging rights is more "hey, I have more computers working and/or dedicated to the purpose of this project than you."...which in turn maybe promotes people to try and "beat" me. In the end, the project wins because people become competitive...building teams, dedicating more computer time, dedicating more computers, etc. -Eric
|
|
Sid Celery Send message Joined: 11 Feb 08 Posts: 2130 Credit: 41,424,155 RAC: 14,205
|
So... if credits are not important for some of you... what alternative you have to measure the computational power working for R@H (or any other BOINC project)? It's none of my business to measure the computational power working for R@H. Why would I want to know? I also don't make the mistake of equating computational power with the value of the work that's done. Maybe it's the same, maybe not, but they're unrelated (unless the credit system is about a million times cleverer than I think it is). That said I think it's pretty reasonable to contextualise a computer's contribution for people and numbers are a pretty good way of doing that - in much the same way I flick through my SO's magazines every so often. Both look kind of nice. Mikey has it about right - I guess I'd want to know if it was worth my while adding an 80486 machine or not (probably not) and if I was upgrading and had an option of a free (or nearly free) add-on or one type of graphics card instead of another the numbers help to see the value of that. As far as whether one project is better or more worthwhile than another I read what's available and I go with it or not. I'm in no position to determine the direction or real value of what's done and it makes no great difference if I'm getting twice as may points as you or only one tenth. I can only do what I can.
|
|
P . P . L . Send message Joined: 20 Aug 06 Posts: 581 Credit: 4,865,274 RAC: 0 |
Hi all. I think this is the highest i've seen it! TeraFLOPS estimate: 106.918 Way to go. 
|
|
svincent Send message Joined: 30 Dec 05 Posts: 219 Credit: 12,120,035 RAC: 0 |
Einstein@home has been down for nearly a week. I imagine this accounts for part of the increase. |
Greg_BE Send message Joined: 30 May 06 Posts: 5691 Credit: 5,859,226 RAC: 0 |
Their fileserver went deep south. It may be back later this week, wait and see what happens to the teraflops after its back up. |
Message boards :
Number crunching :
How many people are crunching Rosetta?
©2024 University of Washington
https://www.bakerlab.org