Can't figure why they need the results of thousands of WUs (with 4 copies of each) within some particular timeframe, so 14 days seems reasonable...unless the developers post a reason to the contrary.
And yes, the estimated time should be increased. I did a survey on seti and found estimated vs actual completion times for Einstein (very short survey, hearsay). Ratio avg was 0.89.
By way of comparison Seti's was 1.63 and Predictor's was 2.16.
(Ratio = Predicted / Actual)
So raising rsc_fpops_est to 61e12 vs the current 40e12 would give somewhat better results.
----
Also there is a factor many new hosts might not be aware of...
If you've just installed BOINC, it has no way of guessing how long you leave the machine on every week. Its initial value is 100% or 24/7. As time progresses boinc updates its values of %time on vs off, and eventually gets accurate ammount.
Until then, if you tell it to connect every ...say 5 days...it will request 5 days of 24 hour work (modified by your resource share).
Ben you´re absolutly right.
But what increases the problem is that when i set up my work for 7 days because i´m on holiday i will get only 6 days from seti but 10 days from einstein.
Thats why the time estimate are oposite on this two projects.
For example seti estimate for one WU 4.05 houres it took around 3.15 what is 50 minutes less.
On einstein the manager estimates 5.51 but actual needs 6.10.
So the einstein sheduler asked for more work that couldn´t finished in time.
> > his never ending insistance that if he were accomdated that world would
> > be safe for democracy.
>
> If that is not mocking, please point out where I said the world would be safe
> etc.
>
> And never ending? Come on. Ive made a total of 12 posts in this forum to
> date.
Yeah, I guess you were right. I did say that ... probably should not have ...
But I did have another thought. You should be using CPDN as your example as they have a year long deadling ...
But, that is somewhat mocking too so I guess it really is time to quit. Done.
> > his never ending insistance that if he were accomdated that world would
> > be safe for democracy.
>
> If that is not mocking, please point out where I said the world would be safe
> etc.
>
> And never ending? Come on. Ive made a total of 12 posts in this forum to
> date.
Yeah, I guess you were right. I did say that ... probably should not have ...
But I did have another thought. You should be using CPDN as your example as they have a year long deadling ...
But, that is somewhat mocking too so I guess it really is time to quit. Done.
> It is funny that someone said the Pentium II computer wouldn't be able to run
> anything above Need For Speed 2, because I have Need For Speed 3, and it works
> great even when the graphics are set to best quality. A lot of my games have
> higher requirements than Need For Speed 3, and they all work fine. SimCity
> 3000 was a little slow, but I installed some more RAM, and now that works
> great. Some of my games won't install on the computer because it says my
> computer is too slow, but when I manually copy the files from the CD, they all
> run at a good speed. Counter-Strike and Half Life run smoothly on my Pentium
> II.
>
Notice I said Half Life 2 NOT Half Life, but yes, you're right about the purpose of the thread!
A very good point made about the deadlines though... This is what I've been trying to say all along!!
I just don't understand why they are in such a hurry to get the results back. Like I've been saying, it takes a while to complete a work unit, but my computer is well within the requirements to run the program. I have 27 GB of free hard disk space and I have 384 MB of RAM. That may not be very much RAM, but I rarely use more than 200 MB. How can anyone say the computer is too old for this project? I have my new computer working on a cancer research grid project, which I feel is more important. You may disagree, but I don't plan on changing my mind about that. So, my Pentium II is the only way I can contribute to this project, which I still find to be interesting even though it isn't essential for people to live.
The problem is not with Einstein's deadline, but lies within Boinc itself. The scheduler that handles the swapping of projects and the controller of the cache needs improvement. Much discussion of the changes needed and the requirements that need to be met for various projects and computers and connection styles have bounced between the members of the Boinc alpha mailing list. Once the finalized parameters were agreed upon, one member has undertaken to program a simulator showing how the new scheduler will work. We are awaiting his results, if it meets are the requirements set down for it, the changes will be forwarded to the Boinc developement team for their approval and implementation. The changes will also require some changes to the server side code of Boinc also, thus requiring all projects to upgrade. The first couple of schedulers and cache controllers that Boinc has grown up with were ok, when there was just a few projects. But now that the number of projects has increased, their limitations have become rapidly apparent. So please bear with it, things will improve.
> The problem is not with Einstein's deadline, but lies within Boinc itself. The
> scheduler that handles the swapping of projects and the controller of the
> cache needs improvement. Much discussion of the changes needed and the
> requirements that need to be met for various projects and computers and
> connection styles have bounced between the members of the Boinc alpha mailing
> list. Once the finalized parameters were agreed upon, one member has
> undertaken to program a simulator showing how the new scheduler will work. We
> are awaiting his results, if it meets are the requirements set down for it,
> the changes will be forwarded to the Boinc developement team for their
> approval and implementation. The changes will also require some changes to the
> server side code of Boinc also, thus requiring all projects to upgrade. The
> first couple of schedulers and cache controllers that Boinc has grown up with
> were ok, when there was just a few projects. But now that the number of
> projects has increased, their limitations have become rapidly apparent. So
> please bear with it, things will improve.
>
It was easier for me to modify the code, and see what it does in actual fact. There are two or three of us running it, and it appears to work ok. Now to get some time from Berkeley for integration...
These last two posts was what I was hoping for when I participated in this thread in the first place... It's nice to have some official input from the project devs into issues like this as this helps put to rest some of the conflicting views around here.
Though one question still remains in my mind, and I would like one of the devs to explain it here, or somewhere on this site for all participants to see, and that is the question... Why are the deadlines only a week? What is the reason for the week cut off?
If some dev could take five minutes to explain this simply, as part of the paragraph on the front page explaining the nature of the experiment, I wouldn't be quite so cut up about missing these deadlines. Although the information present is adequate for the new user, it lacks some of the answers to the more technical questions like that of how a longer deadline would be detrimental to the project and what exactly happens to a validated result back at base? What information is actually used? How is it used? This would be nice to have presented to us in a similar way to how S@H goes into it for those of us who are interested in an in-depth look at what's going on behind the scenes.
I, like many people here by the look of it, just don't understand why there is such a rushed requirement to get results back. Maybe if we could have some of the behind the scenes stuff, then these things would become more apparent... I have to agree with Josh there.
> I, like many people here by the look of it, just don't understand why there is
> such a rushed requirement to get results back. Maybe if we could have some of
> the behind the scenes stuff, then these things would become more apparent... I
> have to agree with Josh there.
Granted I am not an official developer.
But, the official reason is to keep the data population on the database down.
I, and others, have been saying this. What is so hard to understand?
If you want, search for Bruce Allen in the forums, if you go back a couple months he said just that very thing ... and if you look close at the copyright you cannot get much more official than that ...
My opinion: Can't figure
)
My opinion:
Can't figure why they need the results of thousands of WUs (with 4 copies of each) within some particular timeframe, so 14 days seems reasonable...unless the developers post a reason to the contrary.
And yes, the estimated time should be increased. I did a survey on seti and found estimated vs actual completion times for Einstein (very short survey, hearsay). Ratio avg was 0.89.
By way of comparison Seti's was 1.63 and Predictor's was 2.16.
(Ratio = Predicted / Actual)
So raising rsc_fpops_est to 61e12 vs the current 40e12 would give somewhat better results.
----
Also there is a factor many new hosts might not be aware of...
If you've just installed BOINC, it has no way of guessing how long you leave the machine on every week. Its initial value is 100% or 24/7. As time progresses boinc updates its values of %time on vs off, and eventually gets accurate ammount.
Until then, if you tell it to connect every ...say 5 days...it will request 5 days of 24 hour work (modified by your resource share).
Hi Ben you´re absolutly
)
Hi
Ben you´re absolutly right.
But what increases the problem is that when i set up my work for 7 days because i´m on holiday i will get only 6 days from seti but 10 days from einstein.
Thats why the time estimate are oposite on this two projects.
For example seti estimate for one WU 4.05 houres it took around 3.15 what is 50 minutes less.
On einstein the manager estimates 5.51 but actual needs 6.10.
So the einstein sheduler asked for more work that couldn´t finished in time.
greet Mike
> > his never ending
)
> > his never ending insistance that if he were accomdated that world would
> > be safe for democracy.
>
> If that is not mocking, please point out where I said the world would be safe
> etc.
>
> And never ending? Come on. Ive made a total of 12 posts in this forum to
> date.
Yeah, I guess you were right. I did say that ... probably should not have ...
But I did have another thought. You should be using CPDN as your example as they have a year long deadling ...
But, that is somewhat mocking too so I guess it really is time to quit. Done.
> > his never ending
)
> > his never ending insistance that if he were accomdated that world would
> > be safe for democracy.
>
> If that is not mocking, please point out where I said the world would be safe
> etc.
>
> And never ending? Come on. Ive made a total of 12 posts in this forum to
> date.
Yeah, I guess you were right. I did say that ... probably should not have ...
But I did have another thought. You should be using CPDN as your example as they have a year long deadling ...
But, that is somewhat mocking too so I guess it really is time to quit. Done.
> It is funny that someone
)
> It is funny that someone said the Pentium II computer wouldn't be able to run
> anything above Need For Speed 2, because I have Need For Speed 3, and it works
> great even when the graphics are set to best quality. A lot of my games have
> higher requirements than Need For Speed 3, and they all work fine. SimCity
> 3000 was a little slow, but I installed some more RAM, and now that works
> great. Some of my games won't install on the computer because it says my
> computer is too slow, but when I manually copy the files from the CD, they all
> run at a good speed. Counter-Strike and Half Life run smoothly on my Pentium
> II.
>
Notice I said Half Life 2 NOT Half Life, but yes, you're right about the purpose of the thread!
A very good point made about the deadlines though... This is what I've been trying to say all along!!
I just don't understand why
)
I just don't understand why they are in such a hurry to get the results back. Like I've been saying, it takes a while to complete a work unit, but my computer is well within the requirements to run the program. I have 27 GB of free hard disk space and I have 384 MB of RAM. That may not be very much RAM, but I rarely use more than 200 MB. How can anyone say the computer is too old for this project? I have my new computer working on a cancer research grid project, which I feel is more important. You may disagree, but I don't plan on changing my mind about that. So, my Pentium II is the only way I can contribute to this project, which I still find to be interesting even though it isn't essential for people to live.
The problem is not with
)
The problem is not with Einstein's deadline, but lies within Boinc itself. The scheduler that handles the swapping of projects and the controller of the cache needs improvement. Much discussion of the changes needed and the requirements that need to be met for various projects and computers and connection styles have bounced between the members of the Boinc alpha mailing list. Once the finalized parameters were agreed upon, one member has undertaken to program a simulator showing how the new scheduler will work. We are awaiting his results, if it meets are the requirements set down for it, the changes will be forwarded to the Boinc developement team for their approval and implementation. The changes will also require some changes to the server side code of Boinc also, thus requiring all projects to upgrade. The first couple of schedulers and cache controllers that Boinc has grown up with were ok, when there was just a few projects. But now that the number of projects has increased, their limitations have become rapidly apparent. So please bear with it, things will improve.
> The problem is not with
)
> The problem is not with Einstein's deadline, but lies within Boinc itself. The
> scheduler that handles the swapping of projects and the controller of the
> cache needs improvement. Much discussion of the changes needed and the
> requirements that need to be met for various projects and computers and
> connection styles have bounced between the members of the Boinc alpha mailing
> list. Once the finalized parameters were agreed upon, one member has
> undertaken to program a simulator showing how the new scheduler will work. We
> are awaiting his results, if it meets are the requirements set down for it,
> the changes will be forwarded to the Boinc developement team for their
> approval and implementation. The changes will also require some changes to the
> server side code of Boinc also, thus requiring all projects to upgrade. The
> first couple of schedulers and cache controllers that Boinc has grown up with
> were ok, when there was just a few projects. But now that the number of
> projects has increased, their limitations have become rapidly apparent. So
> please bear with it, things will improve.
>
It was easier for me to modify the code, and see what it does in actual fact. There are two or three of us running it, and it appears to work ok. Now to get some time from Berkeley for integration...
BOINC WIKI
These last two posts was what
)
These last two posts was what I was hoping for when I participated in this thread in the first place... It's nice to have some official input from the project devs into issues like this as this helps put to rest some of the conflicting views around here.
Though one question still remains in my mind, and I would like one of the devs to explain it here, or somewhere on this site for all participants to see, and that is the question... Why are the deadlines only a week? What is the reason for the week cut off?
If some dev could take five minutes to explain this simply, as part of the paragraph on the front page explaining the nature of the experiment, I wouldn't be quite so cut up about missing these deadlines. Although the information present is adequate for the new user, it lacks some of the answers to the more technical questions like that of how a longer deadline would be detrimental to the project and what exactly happens to a validated result back at base? What information is actually used? How is it used? This would be nice to have presented to us in a similar way to how S@H goes into it for those of us who are interested in an in-depth look at what's going on behind the scenes.
I, like many people here by the look of it, just don't understand why there is such a rushed requirement to get results back. Maybe if we could have some of the behind the scenes stuff, then these things would become more apparent... I have to agree with Josh there.
> I, like many people here by
)
> I, like many people here by the look of it, just don't understand why there is
> such a rushed requirement to get results back. Maybe if we could have some of
> the behind the scenes stuff, then these things would become more apparent... I
> have to agree with Josh there.
Granted I am not an official developer.
But, the official reason is to keep the data population on the database down.
I, and others, have been saying this. What is so hard to understand?
If you want, search for Bruce Allen in the forums, if you go back a couple months he said just that very thing ... and if you look close at the copyright you cannot get much more official than that ...