Classic case of round hole square peg, if I have ever seen one.........Or maybe it is the rectangular hole and the triangular peg, one never knows, I am a cheerleader for BOINC, and I am proud of that fact.......
yes, condemned out of his own mouth
- tho maybe I'd change one of the vowels ;-)
It is worth adding that on SETI he would not be allowed to post until he had returned some work. His attitude is living proof of the motivation for that approach. In fairness tho, he is doing some for us - he has four wu running, and on a decent machine - better than mine. [edit] which he has hidden since I last looked
I'd be sad to see us follow SETI on that - we also get people who do 'post nice' as newcomers, and I would not want to turn them away because of a few hexacontagons.
If my memory serves me correctly you posted something some weeks ago about that the 4.4x client have a limit for how many projects at a time it will hold WU's in the cache from....
just popped in to say that upgrading from 4.25 to 4.5 has solved the majority of workunit not finished on time issues - cheers :)
yes, condemned out of his own mouth
- tho maybe I'd change one of the vowels ;-)
It is worth adding that on SETI he would not be allowed to post until he had returned some work. His attitude is living proof of the motivation for that approach. In fairness tho, he is doing some for us - he has four wu running, and on a decent machine - better than mine. [edit] which he has hidden since I last looked
Quote:
Are you a stalker?
I'd be sad to see us follow SETI on that - we also get people who do 'post nice' as newcomers, and I would not want to turn them away because of a few hexacontagons.
Wow look at me...I have work on e&h so now I can post here. So now I have enough status to put up as much BS as gravy. Of course I actually will work on more then one project and continue adding credit to each...
I just saw a result come in four day overdue and get credit for it. So you don't absolutely have to finish in 7 days. You just need to beat whomever gets the reassigned wu. Should that person get credit even though his result is no longer needed?
I just saw a result come in four day overdue and get credit for it. So you don't absolutely have to finish in 7 days. You just need to beat whomever gets the reassigned wu. Should that person get credit even though his result is no longer needed?
If you do the work you should get the credit if finished within the deadline. If you can't get it done then don't sign up for the project. That just means someone is probably running a dead wu. Just don't like working only for one project because of deadline/size of wu.
Like I said, seems pretty stupid to have such a large wu needing tons of number crunching and setting a short deadline because you have a database problem. No one else has a database problem with the smaller wu's.
Still I find it funny to get credit for something you didn't make deadline on. Wonder if they set it that way?
cegoth:
I agree wholly with your 1st and 3rd paragraphs.
I disagree with your 2nd paragraph. Months of watching my result leads me to conclude that e@h works well. It is rare to see a result remain pending for even a week. It is also rare to see sever or sw issues wasting the cpu cycles I attempt to contribute. Obviously it would be better if the project could make use of installation with fewer resources. But the people who manage and run this project don't have limitless funds or personal resources. I think they have done a fine job with what they have.
cegoth:
I agree wholly with you 1st and 3rd paragraphs.
I disagree with your 2nd paragraph. Months of watching my result leads me to conclude that e@h works well. It is rare to see a result remain pending for even a week. It is also rare to see sever or sw issues wasting the cpu cycles I attempt to contribute. Obviously it would be better if the project could make use of installation with fewer resources. But the people who manage and run this project don't have limitless funds or personal resources. I think they have done a fine job with what they have.
My thought is that I run through tons of seti work. Although theirs does not seem to take the amount of number crunching that e&h does it seems like the wu's are running through faster mainly because I am more likely to do their work rather then e&h. However having said that I am trying something to see what the difference is between different projects. Something tells me that having three machines with exactly the same load, software and hardware wise (and this is down to being exactly the same as far as ever thing possible within my power - I can't balance and blueprint cpu's and memory like a 327's cylinders and valves), will give me a definite result as to who is using boinc to its full potential.
I have read that e&h has three stages, first two run the data, "two different sections (different time ranges and/or detectors) of Gravitational Wave Detector data are searched for the same set of candidate physical sources", and the third compares. Wouldn't it make sense to break up the first part? Basically it sounds like two pieces of data that can be run in two different strings and then compared either at a later date or at the end. Think about a computer that only got the last part to hash. It would be flying through wu's although very small sized and credit wise. It seems they send me work in pairs. My hyperthread would be perfect for running the first two parts together but it's not utilized. I still run two sets of data twice and then compare both at the end. Give me the first two sets of data, when they finish send them back for comparison or let my machine compare and send back. The result comes quicker for each data set and the result is sent back quicker. Less credit but it sounds better. If I lose one the program should restart it and store the other locally. When the other finishes its compared. Meanwhile I have started on another or one thread is working on seti or predictor or whatever til the broken one is redone. This also would STILL require a longer deadline.
It seems they want to get the data as fast as possible by setting short deadlines and large wu's...yet I find it is probably going against them since most people see the needed computation time and then probably set seti or predictor to wail away since they want to see a result. This may be volunteer but human nature says its a competition and people want to see quick results. I bet predictor gives them that and therefore they steer clear of e&h (and cpdn).
cegoth:
What works for SETI will not work for everything else. Different problems require different solutions. SETI is the grandfather of all BOINC projects but there is no reason to presume that all of its decedents will look or feel like SETI. I think it would be a bad idea to expect them to. The descendents of SETI should and hopefully will succeed or fail based on the individual merits not on some preconceived standard.
The fact that the first data set (S3 science run) has been completed and the second (S4) has been started implies at a minimum the people running the project believe it is not failing.
If I had several months to study the design of the e@h project I might presume to tell its authors how to improve their methods and i will not criticize your suggestion. I will pass along the observation that breaking the process into smaller steps may well increase the data management requirements dramatically, it will require writing an application that take different inputs and create different outputs.
RE: Classic case of round
)
With e&h its round hole hexacontagon peg.
Poor gravy...you know I'm right.
RE: With e&h its round
)
is sweet way all insults refer to itself
yes: hexacontagon peg not square one
all agree yes?
RE: RE: With e&h its
)
yes, condemned out of his own mouth
- tho maybe I'd change one of the vowels ;-)
It is worth adding that on SETI he would not be allowed to post until he had returned some work. His attitude is living proof of the motivation for that approach. In fairness tho, he is doing some for us - he has four wu running, and on a decent machine - better than mine. [edit] which he has hidden since I last looked
I'd be sad to see us follow SETI on that - we also get people who do 'post nice' as newcomers, and I would not want to turn them away because of a few hexacontagons.
~~gravywavy
RE: RE: If my memory
)
RE: RE: is sweet way all
)
Wow look at me...I have work on e&h so now I can post here. So now I have enough status to put up as much BS as gravy. Of course I actually will work on more then one project and continue adding credit to each...
I just saw a result come in
)
I just saw a result come in four day overdue and get credit for it. So you don't absolutely have to finish in 7 days. You just need to beat whomever gets the reassigned wu. Should that person get credit even though his result is no longer needed?
RE: I just saw a result
)
If you do the work you should get the credit if finished within the deadline. If you can't get it done then don't sign up for the project. That just means someone is probably running a dead wu. Just don't like working only for one project because of deadline/size of wu.
Like I said, seems pretty stupid to have such a large wu needing tons of number crunching and setting a short deadline because you have a database problem. No one else has a database problem with the smaller wu's.
Still I find it funny to get credit for something you didn't make deadline on. Wonder if they set it that way?
cegoth: I agree wholly with
)
cegoth:
I agree wholly with your 1st and 3rd paragraphs.
I disagree with your 2nd paragraph. Months of watching my result leads me to conclude that e@h works well. It is rare to see a result remain pending for even a week. It is also rare to see sever or sw issues wasting the cpu cycles I attempt to contribute. Obviously it would be better if the project could make use of installation with fewer resources. But the people who manage and run this project don't have limitless funds or personal resources. I think they have done a fine job with what they have.
RE: cegoth: I agree wholly
)
My thought is that I run through tons of seti work. Although theirs does not seem to take the amount of number crunching that e&h does it seems like the wu's are running through faster mainly because I am more likely to do their work rather then e&h. However having said that I am trying something to see what the difference is between different projects. Something tells me that having three machines with exactly the same load, software and hardware wise (and this is down to being exactly the same as far as ever thing possible within my power - I can't balance and blueprint cpu's and memory like a 327's cylinders and valves), will give me a definite result as to who is using boinc to its full potential.
I have read that e&h has three stages, first two run the data, "two different sections (different time ranges and/or detectors) of Gravitational Wave Detector data are searched for the same set of candidate physical sources", and the third compares. Wouldn't it make sense to break up the first part? Basically it sounds like two pieces of data that can be run in two different strings and then compared either at a later date or at the end. Think about a computer that only got the last part to hash. It would be flying through wu's although very small sized and credit wise. It seems they send me work in pairs. My hyperthread would be perfect for running the first two parts together but it's not utilized. I still run two sets of data twice and then compare both at the end. Give me the first two sets of data, when they finish send them back for comparison or let my machine compare and send back. The result comes quicker for each data set and the result is sent back quicker. Less credit but it sounds better. If I lose one the program should restart it and store the other locally. When the other finishes its compared. Meanwhile I have started on another or one thread is working on seti or predictor or whatever til the broken one is redone. This also would STILL require a longer deadline.
It seems they want to get the data as fast as possible by setting short deadlines and large wu's...yet I find it is probably going against them since most people see the needed computation time and then probably set seti or predictor to wail away since they want to see a result. This may be volunteer but human nature says its a competition and people want to see quick results. I bet predictor gives them that and therefore they steer clear of e&h (and cpdn).
cegoth: What works for SETI
)
cegoth:
What works for SETI will not work for everything else. Different problems require different solutions. SETI is the grandfather of all BOINC projects but there is no reason to presume that all of its decedents will look or feel like SETI. I think it would be a bad idea to expect them to. The descendents of SETI should and hopefully will succeed or fail based on the individual merits not on some preconceived standard.
The fact that the first data set (S3 science run) has been completed and the second (S4) has been started implies at a minimum the people running the project believe it is not failing.
If I had several months to study the design of the e@h project I might presume to tell its authors how to improve their methods and i will not criticize your suggestion. I will pass along the observation that breaking the process into smaller steps may well increase the data management requirements dramatically, it will require writing an application that take different inputs and create different outputs.