Is our participation useful?

Stranger7777
Stranger7777
Joined: 17 Mar 05
Posts: 436
Credit: 366,246,183
RAC: 103,050

Let me disagree with both of

Let me disagree with both of you: RandyC and Hoarfrost.

The question I'd asked will give an answer whether our participation useful or not, because it gives us a number - value of participation of ATLAS and us.

When Mike Hewson gives us real numbers or graphs, then we'll see the real picture of this.

2 Hoarfrost: I think you make a mistake when think that your computer box (with video, cooler, hard drive, TV tuner, etc) and only 1 (2 or 4 cores) consumes power lower than the cluster cell that consists of only core (1,2,4,8, etc) and some RAM for it. But it works (when idle of other things) on this project 100% of time, but your system works on Windows (Linux, etc), some other programs, services, other projects and so on. So, it uses only 10-50% of your processor time for real work for this project. And you can see the results of it. Look at TOP 100 users status page. The real measure of our usefulness is the number of cobblestones that we receive. Look at it. And you will see that 2 first users makes more work than thousands of us. But don't hurry to answer me. Let's see what Mike tells us.

hoarfrost
hoarfrost
Joined: 9 Feb 05
Posts: 207
Credit: 51,668,899
RAC: 4,113

RE: Let me disagree with

Message 82481 in response to message 82480

Quote:

Let me disagree with both of you: RandyC and Hoarfrost.

The question I'd asked will give an answer whether our participation useful or not, because it gives us a number - value of participation of ATLAS and us.

When Mike Hewson gives us real numbers or graphs, then we'll see the real picture of this.

2 Hoarfrost: I think you make a mistake when think that your computer box (with video, cooler, hard drive, TV tuner, etc) and only 1 (2 or 4 cores) consumes power lower than the cluster cell that consists of only core (1,2,4,8, etc) and some RAM for it. But it works (when idle of other things) on this project 100% of time, but your system works on Windows (Linux, etc), some other programs, services, other projects and so on. So, it uses only 10-50% of your processor time for real work for this project. And you can see the results of it. Look at TOP 100 users status page. The real measure of our usefulness is the number of cobblestones that we receive. Look at it. And you will see that 2 first users makes more work than thousands of us. But don't hurry to answer me. Let's see what Mike tells us.


I believe in numbers only. :)

1. TV-tuner, Video card and another "non-calculation" elements need for me because my computer need for me as my tool. It works does without any dependence of calculations. And in case of workstation we must take only "addition" energy consumption:

dE = Ecalc - Eusual work

that smaller than full energy, that need for cluster node.
When I sleep, of course dE = Ecalc, but in this time graphics and anoother devices does not works really.

2. Two my cores ~ 3000 CS/day. ALTAS have ~ 5000 cores.
3000 CS/day * 2500 = 7500000 CS/day.
Maximum ATLAS day credit (that I remeber) ~ 5000000 CS, that near to maximum credit of this group of cores without oveclock.
Consequently, ALTAS is an equivalent ONLY 2500 participants as I.

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6,125
Credit: 127,324,678
RAC: 11,673

Whoa guys! :-) I didn't mean

Whoa guys! :-)
I didn't mean to start a fire, or put one out either, I just wanted to define why it is that people contribute here at E@H. This is a recurrent theme, and for me a fascinating one, essentially as to why those who would otherwise be total strangers join forces to solve a problem. As the matter came around again, as it does every 3 to 6 months, I thought I'd ask. I wasn't constructing some sideways aimed proposal - what you read is what I meant! :-)

Specifics:

- the relative merit of the contributors is already there in the participants list. In an idle moment or three, I just summated in ever widening bands to see what happened and found a power law. You could make the geometric analogy of a pyramid - but one that reforms itself perpetually. And yes, without supercomputing clusters it is a lesser one, but ditto with volunteers. Mathematically I successively integrated longer stretches from the number one spot down. You take differences ( differentiate ) to get from my figures back again. RAC is a weighted rolling average of credit and doesn't mean quite the same as credit/time but is a reasonable proxy for that and has 'decay'.

- as for electricity I was demonstrating how the management had said 'thank you' for such contributions. Actually somewhere downstream from that post I quoted was a discussion of the relative costs of electrical supply around the world. I can't find the reference. I think in particular it was revealed to be then especially expensive in eastern Europe ( old Soviet block countries ) and rather cheaper in some parts of the US. Discussion also mentioned the cost of running air-conditioning in summertime and how that might affect participation then. I have made a few jokes from time to time about having heaters as calculators, so I don't mind them running in winter ( I live half way up a mountain range ). My study is always the warmest room in the house. We even wondered whether other household appliances could contribute, "Toasters For Einstein" and that sort of thing. :-)

- in any case there are genuine distinctions to be made between computers which were bought and constructed purely for E@H purposes, computers 'leased' for E@H time, computers paid for publicly vs privately, computers with a non-BOINC justification for their existence doing E@H 'on the side', and computers BOINC'ing with E@H as one of several projects. The point has also been made several times that if you don't own the box then you must ask the one who does before committing their resources - this now including allied costs like connectivity, maintenance, possible distraction from a business task etc. So E@H does not condone this type of theft ( yes, it is theft if it is not yours and you didn't get permission ). There are no doubt other categories.

- there are mixes of these 'types' all around. Because I'm a nerd ( as are others ) I have an E@H farm of boxes, doing nearly pure E@H work - but also serving as file repositories, decision support roles ( drug databases for prescribing ), software development platforms, games boxes ( ie. my kids who I con into contributing by bribing them ) and just old favourites I can't bear to turn off. My accountant would no doubt recommend the tossing of most on the nature strip to await the next hard rubbish collection by the local council. But, like doctors, you shouldn't always do what they say! :-)

- so such role mixing can fairly trash any easy assessment of value/cost in real world terms. So I think it is hard to assume ( and I not saying anyone is mind you ) that contributors actually do such a thing. They probably make a largely non-economic value judgment, propelled by some deeper or even unrelated reason. It might just be the Einstein name, for instance.

- so I was sort of getting at that area with those questions I posed. Meaning that were people actually heavily 'rating' their contribution against others and perhaps lining their motives up against other peoples motives? How much does it matter to some of us whether or not X or Y or Z is also contributing? Do people do it for some herd instinct to congregate ( like a footy club ), to compete ( also like a footy club ), or to just enjoy the activity ( yes, like a footy club too ). There are a wealth of BOINC related teams out there.

Either way, don't take me too seriously ... :-)

Cheers, Mike.

I have made this letter longer than usual because I lack the time to make it shorter. Blaise Pascal

tullio
tullio
Joined: 22 Jan 05
Posts: 2,043
Credit: 39,991,044
RAC: 9,114

As a contributor to

As a contributor to climateprediction.net and CPDN Beta of the UK Meteorological Office (very big programs of millions of FORTRAN code lines) I have been asked by an Oxford graduate student working towards his PhD to take part in a survey researching the motives that make people donate their computer times and electricity bills to volunteer distributed computing. The climate simulation models run on very extended deadlines. I am working on a task that extends its deadline to November 2010. This is just an extreme example. I am looking forward to read the student's PhD thesis.
Tullio
As far as I understand these models, they consist in solving an equation or a set of equations. Each user is given a set of slightly different initial conditions and the final results are then compared. The more users and the higher the number of final results to be compared.

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6,125
Credit: 127,324,678
RAC: 11,673

RE: As a contributor to

Message 82484 in response to message 82483

Quote:
As a contributor to climateprediction.net and CPDN Beta of the UK Meteorological Office (very big programs of millions of FORTRAN code lines) I have been asked by an Oxford graduate student working towards his PhD to take part in a survey researching the motives that make people donate their computer times and electricity bills to volunteer distributed computing. The climate simulation models run on very extended deadlines. I am working on a task that extends its deadline to November 2010. This is just an extreme example. I am looking forward to read the student's PhD thesis.
Tullio
As far as I understand these models, they consist in solving an equation or a set of equations. Each user is given a set of slightly different initial conditions and the final results are then compared. The more users and the higher the number of final results to be compared.


Neat, what is the stated field of the PhD? Sociology, psychology, economics, computing ... ?

I might add to my last post - yes supercomputing clusters will displace quite a few contributors who are in lower rankings. Probably some large ratio. That's what the 'super' bit means I guess. If I'd paid for such a cluster myself I'd be disappointed if it didn't perform that way. The presence of such clusters in an otherwise volunteer effort could be a basis for some of said volunteers withdrawing ( ie. what is the use? as per the thread title ) - after all you can supply or deny on any basis what-so-ever. I'd like to hear the specific reasoning for that. Mind you those that have left E@H aren't likely to be reading or responding here. It is reminiscent of voter participation arguments in 'democracies'.

Such a bland swapping of computing power however is still an orthogonal quality to the other aspects of distributed contributing. Volunteering, pretty well by definition, is an act that bypasses a pure economic/efficiency analysis - there's no personal monetary profit in it. So what makes for the interest is what are the non-economic motivators. :-)

Cheers, Mike.

I have made this letter longer than usual because I lack the time to make it shorter. Blaise Pascal

Stranger7777
Stranger7777
Joined: 17 Mar 05
Posts: 436
Credit: 366,246,183
RAC: 103,050

Yes, you're absolutely

Yes, you're absolutely right.

But your answer again doesn't answer the theme of the thread in numbers.

As far as I can see in the user statistics, it is better to have ATLAS working when idle on E@H then to have work with at least ~25000 volunteers (to look from the end of user-list) over the world: generate Wus that will some time never return, send them out several times (consume traffic, time and electricity for this) and wait for a deadline to understand that it should be resent again.
Instead, ATLAS is working locally (no internet traffic needed) and need no such a lot of extended maintanance (it is very stable) and it already consumes electricity, so this is already prepaid.

So, yes, the participation is useful... but for those of us who contribute a valuable amount of work or contribute knowledge and research time in this field of science. Others, unfortunelly, if will stop work one day toghether will not make a sense to the whole project run.

May be I'm mistaking?

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6,125
Credit: 127,324,678
RAC: 11,673

RE: Yes, you're absolutely

Message 82487 in response to message 82485

Quote:

Yes, you're absolutely right.

But your answer again doesn't answer the theme of the thread in numbers.

As far as I can see in the user statistics, it is better to have ATLAS working when idle on E@H then to have work with at least ~25000 volunteers (to look from the end of user-list) over the world: generate Wus that will some time never return, send them out several times (consume traffic, time and electricity for this) and wait for a deadline to understand that it should be resent again.
Instead, ATLAS is working locally (no internet traffic needed) and need no such a lot of extended maintanance (it is very stable) and it already consumes electricity, so this is already prepaid.

So, yes, the participation is useful... but for those of us who contribute a valuable amount of work or contribute knowledge and research time in this field of science. Others, unfortunelly, if will stop work one day toghether will not make a sense to the whole project run.

May be I'm mistaking?


You are right also! :-)

It's just that one can have different criteria for measuring success or 'usefulness' or whatever. If ATLAS is running flat chat for E@H it will surely 'displace' many thousands I think. Inevitably the physics of time and distance will make that type of arrangement preferable/efficient/economic if measured on those grounds alone.

Quote:
[ I think we had a chuckle recently - at least I thought it was funny 'cos I made the joke - that a network with a star topology using optical fibre would be better for E@H. ]

Distributed computing is where science meets the masses. Even Einstein had something to say about 'common sense' ( being a set of prejudices obtained while growing up ). One can be surprised by peoples motives. From the project's point of view any contribution is a good one, regardless of whatever specific motivations float particular boats - the science is getting done regardless.

If people want to say 'why bother with this because of some supercomputer X that will do my WU's in Y% of the time anyway' that's their call. I just can't see the difference between that and 'why bother because some volunteer Z will do my WU's in Q% of the time anyway'. There is a technical disparity as you outline, but in my experience, based on observation, is that those who express themselves at the boards don't do that calculation much at all.

These are generalities of course. Mind you there is a much vaster set of lurkers out there of whom we'll never know in detail of. Conceivably there is the greater fraction of users who neither know nor care about this board - not that they must or should - and simply BOINC away trusting such concerns to others.

Cheers, Mike.

NB. I'll delete your double post for you.

I have made this letter longer than usual because I lack the time to make it shorter. Blaise Pascal

Stranger7777
Stranger7777
Joined: 17 Mar 05
Posts: 436
Credit: 366,246,183
RAC: 103,050

Thanks for moderation. I'll

Thanks for moderation. I'll try to be carefull next time :)

I suspect that the participation here is a kind of democracy trend: everyone could be a participant and all have the posibilities to grow on the crunch scene, either on sience part or on crunchers list. And that's make us unite in front of horizon of discaveries.

But, nevertheless, it will be very interesting not only to understand the skygrid statistics, but to visualize the user statistics in way we are talking about. Just to see, how can we divide users into groups for e.g. with equal amount of work done (moving apart supercomputers at all). Or to see the percent of groups that work only by one or two members, whos work overcomes the power of the others. It will be very interesting to know how many really active users are working on the project, instead of that found on stats page. We can take for criteria the amount work > 0.01% of the overall productivity and so on.

This numbers and graphs can really help project organizers to see the work statistics that they will never work on because of lack of time. And it will help us to see, should we advertise more people to the project or get new and old rigs ourselves on a board instead of this.

tullio
tullio
Joined: 22 Jan 05
Posts: 2,043
Credit: 39,991,044
RAC: 9,114

RE: Neat, what is the

Message 82489 in response to message 82484

Quote:


Neat, what is the stated field of the PhD? Sociology, psychology, economics, computing ... ?

I might add to my last post - yes supercomputing clusters will displace quite a few contributors who are in lower rankings. Probably some large ratio. That's what the 'super' bit means I guess. If I'd paid for such a cluster myself I'd be disappointed if it didn't perform that way. The presence of such clusters in an otherwise volunteer effort could be a basis for some of said volunteers withdrawing ( ie. what is the use? as per the thread title ) - after all you can supply or deny on any basis what-so-ever. I'd like to hear the specific reasoning for that. Mind you those that have left E@H aren't likely to be reading or responding here. It is reminiscent of voter participation arguments in 'democracies'.

Such a bland swapping of computing power however is still an orthogonal quality to the other aspects of distributed contributing. Volunteering, pretty well by definition, is an act that bypasses a pure economic/efficiency analysis - there's no personal monetary profit in it. So what makes for the interest is what are the non-economic motivators. :-)

Cheers, Mike.


I don't know, but he promised to write me again and I shall ask him. As far as I can know, psychology might be the answer. People tend to enlist in projects like folding@home which promise to find a cure for cancer, AIDS and the like. SETI was a starting point, and everything else kind of piggy-backed on its success. Einstein@home is more away from common people's interests and yet it has a good number of followers. Maybe people are not satisfied with the ivory tower mentality of many scientists and are willing to take part in a research.Astronomy always had the support of amateur astronomers. It is more difficult to be an amateur physicist and amateur chemists have burned many homes with their experiments.
Tullio

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6,125
Credit: 127,324,678
RAC: 11,673

RE: Thanks for

Message 82490 in response to message 82488

Quote:
Thanks for moderation..... I'll try to be carefull next time :)


Dear me, I get to moderate myself if I do a boner ...... but the other mods get to find out though! :-)

Quote:
I suspect that the participation here is a kind of democracy trend: everyone could be a participant and all have the posibilities to grow on the crunch scene, either on sience part or on crunchers list. And that's make us unite in front of horizon of discaveries.

Yep the whole Einstein, spacetime, black holes, neutron star exotica territory has a tremendous mystique. I think the 'gravity wiggles in your own home' has really tickled more than a few of us, regardless of whether we understand even the basics of it.

Quote:
But, nevertheless, it will be very interesting not only to understand the skygrid statistics, but to visualize the user statistics in way we are talking about. Just to see, how can we divide users into groups for e.g. with equal amount of work done (moving apart supercomputers at all). Or to see the percent of groups that work only by one or two members, whos work overcomes the power of the others. It will be very interesting to know how many really active users are working on the project, instead of that found on stats page. We can take for criteria the amount work > 0.01% of the overall productivity and so on.

Agreed, the stats page doesn't tell us enough to quite answer what you're after.

Quote:
This numbers and graphs can really help project organizers to see the work statistics that they will never work on because of lack of time. And it will help us to see, should we advertise more people to the project or get new and old rigs ourselves on a board instead of this.

It'd be nice to distil and bottle the impetus! :-)

Heck, if I had the time, and access to the data, I'd like to do 'cohort lifetimes' ( which I've mentioned before ). Roughly this means sampling the E@H population in a random way ( as it's probably too onerous to analyze everybody in E@H ) to get subsets that match up for various characters like machine performance, RAC, credit, OS, country, Internet access method, when you came to E@H, your dog's brothers' toenail color - whatever. See if there are reliably discernible trends and in particular what correlates with persistent contribution to the project. This could divulge not only the answer to 'why did you volunteer?' but also 'why do you continue to do so?' - by seeing which, if any, characteristics are predictors of longevity. There may of course be none, but I suspect 'emergent' behaviours not foreseen.

Quote:
'Cohort' is a word derived from an old Roman/Latin word referring to a group of trainee Legionnaires that can be followed throughout their military careers. From 'first stab' to death/retirement. It is a concept that is used alot in medical research .... fortunately we no longer 'decimate', that is kill one out of every ten in order to encourage the remaining nine.

Some we already know, or can safely predict. Like that dial-up is the commonest method of ISP access worldwide. Prof Bruce has wanted the up/download strategy of E@H files not to prejudice the involvement of those so constrained. So you wouldn't likely get long term involvement from those forced to 'risk' large downloads over 28K lines. IF one then associates access speed to computing power ( since both will relate to a higher income bracket ) THEN the more productive machines will be having a greater number of ISP access ( 'cos we're keeping the files smaller for the 28K people ) than would be most efficient for their T1 speed channel, say. Overall administrative overhead for a given total data quantity transferred is generally lower with fewer files. With packet based systems ( ie. most of the Internet ) this is significant.

Quote:
We've been promised ASDL 2 locally, Victoria/Australia, which seems to imply a ~10Meg rate. However you'd want to know how many share that fast lane before buying in.

But there's inevitable tradeoff. As discussed here the latter part of the calculation pipeline, of which E@H is part of, can drive the choices as the entire flow must unify. Note that while we may be free to volunteer as per any basis of pleasure, public funding however is usually attached to some sort of deed of arrangement with providers of said funds. The NSF in USA I think. There'll be constraints we haven't imagined there I suppose - plus auditing activities which could tax the most patient of souls.

Cheers, Mike.

I have made this letter longer than usual because I lack the time to make it shorter. Blaise Pascal

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.