BOINC porting

Please confine these topics to things that would be of general interest to those who are interested in FAH which don't fall into any other category.

Moderator: Site Moderators

BOINC porting

Postby cenit » Tue Feb 24, 2009 5:48 pm

only two words: why not?

Starting from v6 BOINC has:
- SMP support

Starting from v6.4 there's:
- full official nVidia CUDA support

Starting from v6.8 (will be out in few months, after final v6.6 that will have a new scheduler):
- full official ATi STREAM support.

It seems to me that it's a perfect time-table to start thinking about porting to that platform, IMHO better than what F@H gives you.
Are you (PandeGroup) worried to loose many clients that will find a whole new world behind F@H??

there was a trial (http://fah-boinc.stanford.edu/), what's happened?

Thanks a lot!
Last edited by cenit on Thu Feb 26, 2009 3:58 pm, edited 2 times in total.
User avatar
cenit
 
Posts: 33
Joined: Sat Dec 27, 2008 3:13 pm

Re: BOINC porting

Postby Ivoshiee » Tue Feb 24, 2009 6:00 pm

There was a FAH BOINC beta some years back, but all development on it stopped. Since then the BOINC developer has left the PG.
Maybe there is some new advances in BOINC that can ease FAH development for it, but that is for the PG to answer.
All current signs do not indicate anything BOINC related and FAH own platform is still being actively developed.
Ivoshiee
Site Moderator
 
Posts: 1286
Joined: Sun Dec 02, 2007 12:05 am
Location: Estonia

Re: BOINC porting

Postby John Naylor » Tue Feb 24, 2009 6:41 pm

The main reason quoted for not moving to BOINC now is that it requires the software package to be made open source... and therefore to protect the integrity of the results each work unit needs to be folded multiple times, to protect against those who try to cheat/modify the client files. To ensure that each unit only needs folding once and that only official versions of the scientific software are used (as the tiniest change can invalidate the results), the client has to be closed source, which therefore means BOINC is not an option. The extra hardware support (or rather, what used to be a lack of it) is not the problem...
Folding whatever I'm sent since March 2006 :) Beta testing since October 2006. www.FAH-Addict.net Administrator since August 2009.
User avatar
John Naylor
 
Posts: 950
Joined: Mon Dec 03, 2007 4:36 pm
Location: University of Birmingham, UK

Re: BOINC porting

Postby Ivoshiee » Tue Feb 24, 2009 6:52 pm

That is not a real reason at all - there is no need to OSS anything and FAH client can be bundled inside BOINC framework (tested).
Ivoshiee
Site Moderator
 
Posts: 1286
Joined: Sun Dec 02, 2007 12:05 am
Location: Estonia

Re: BOINC porting

Postby John Naylor » Tue Feb 24, 2009 7:04 pm

I was under the impression BOINC was LGPL and to become part of it the scientific cores of F@H have to be released under the LGPL too, thereby compromising security... If it was as secure as the current system then SETI would not still require multiple results to validate each other...
User avatar
John Naylor
 
Posts: 950
Joined: Mon Dec 03, 2007 4:36 pm
Location: University of Birmingham, UK

Re: BOINC porting

Postby toTOW » Tue Feb 24, 2009 7:11 pm

I think one of the reason that FAH is not available on BOINC has to do with the rapid turnaround that FAH need.

PandeGroup has always said that the faster the WUs are returned, the faster the project makes progress. BOINC software allows you to share your computing power between various projects, which might slow the FAH turnover. BOINC software makes sure that the deadlines are met, but we all know that in FAH, it's better to return the WU as fast as possible ;)
Folding@Home beta tester since 2002. Folding Forum moderator since July 2008.

FAH-Addict : latest news, tests and reviews about Folding@Home project.

Image
User avatar
toTOW
Site Moderator
 
Posts: 8696
Joined: Sun Dec 02, 2007 10:38 am
Location: Bordeaux, France

Re: BOINC porting

Postby 7im » Tue Feb 24, 2009 7:23 pm

Hello cenit, welcome to the folding support forum.

Why not? is not the right question. WHY? is the better question.

Why would Stanford what to saddle itself with porting backwards to a client that was designed by committee? Too many compromises, and too slow to progress. Fah had stream processing clients more than 3 years ago, and BOINC is just now adding it? FAH is blazing the trail for other distributed computing projects to follow. Fah had the first GPU client. First PS3 client. And has had a true SMP client years ahead of BOINC.

Don't get me wrong, I'm not putting BOINC down as it has its place and purpose. But fah tried this once and didn't have a lot of success.

Have you tried the fah client?
How to provide enough information to get helpful support
Tell me and I forget. Teach me and I remember. Involve me and I learn.
User avatar
7im
 
Posts: 14648
Joined: Thu Nov 29, 2007 4:30 pm
Location: Arizona

Re: BOINC porting

Postby Ivoshiee » Tue Feb 24, 2009 7:42 pm

John Naylor wrote:I was under the impression BOINC was LGPL and to become part of it the scientific cores of F@H have to be released under the LGPL too, thereby compromising security... If it was as secure as the current system then SETI would not still require multiple results to validate each other...

BOINC itself can be under whatever license, but it does not force you to hand over your source for all to see and tinker with.
Ivoshiee
Site Moderator
 
Posts: 1286
Joined: Sun Dec 02, 2007 12:05 am
Location: Estonia

Re: BOINC porting

Postby Beberg » Tue Feb 24, 2009 9:30 pm

BOINC has one very specific model for the types of projects it supports. Folding@home does not fit into that model. That doesn't make it bad, just different and thus incompatible.
User avatar
Beberg
Pande Group Member
 
Posts: 566
Joined: Sat Dec 01, 2007 9:05 pm
Location: Stanford

Re: BOINC porting

Postby cenit » Thu Feb 26, 2009 2:29 pm

John Naylor wrote:I was under the impression BOINC was LGPL and to become part of it the scientific cores of F@H have to be released under the LGPL too, thereby compromising security... If it was as secure as the current system then SETI would not still require multiple results to validate each other...


SETI requires multiple results for comparisons, for code check, for many reasons... who knows? Many projects don't require multiple results (Rosetta is an example), and, most important, you don't have to opensource your code!

7im wrote:Hello cenit, welcome to the folding support forum.
Why not? is not the right question. WHY? is the better question.
Why would Stanford what to saddle itself with porting backwards to a client that was designed by committee? Too many compromises, and too slow to progress. Fah had stream processing clients more than 3 years ago, and BOINC is just now adding it? FAH is blazing the trail for other distributed computing projects to follow. Fah had the first GPU client. First PS3 client. And has had a true SMP client years ahead of BOINC.
Don't get me wrong, I'm not putting BOINC down as it has its place and purpose. But fah tried this once and didn't have a lot of success.
Have you tried the fah client?


Hi 7im.
I've tried the fah client, in fact I give FAH my ATi 4870, at least until a good project under boinc will support it. I've asked this for this reason: could it be that I'm the ONLY ONE which will leave F@H that day? I'm sure I'm not the only one
In fact, I would like to manage all in a single place, it's better if you don't want to support only one project. If I revert the question, have you ever tried boinc? It's really superior to the fah client!
And for the progress on boinc, don't you think that if PandeGroup had submitted patches to boinc 3 years ago to support GPU, everybody would have benefitted from it? BOINC is an open platform and everybody could help writing code! Doing all in this closed-sourced way has only duplicated efforts!
(for the ps3, boinc is only a platform, virtually running everywhere (it runs on the ps3, some projects support it). It stands on the projects to support as many clients as possible)

toTOW wrote:I think one of the reason that FAH is not available on BOINC has to do with the rapid turnaround that FAH need.
PandeGroup has always said that the faster the WUs are returned, the faster the project makes progress. BOINC software allows you to share your computing power between various projects, which might slow the FAH turnover. BOINC software makes sure that the deadlines are met, but we all know that in FAH, it's better to return the WU as fast as possible ;)


This seems the biggest reason to me, but in a different view: I'm sure that Pande Group are feared to loose clients. I know so many people that don't know anything about distributed computing, but 70% of people who donate do D.C. didn't know the existance of BOINC and the projects that run on it. When they know about it, some leave (yes!) and some runs now more projects, leaving FAH ultimately with less resources. But you cannot say FAH is the best project and so they did something stupid. You can't, in any way!

Beberg wrote:BOINC has one very specific model for the types of projects it supports. Folding@home does not fit into that model. That doesn't make it bad, just different and thus incompatible.

You seem to be a Pande Group Member. I think that from your position you can't say so! Would you mean that Einstein@Home, Rosetta, World Community Grid are less important than you? From your words it seems so! How could you think you couldn't fit with them? Do you think that your "model" (??) is superior?
User avatar
cenit
 
Posts: 33
Joined: Sat Dec 27, 2008 3:13 pm

Re: BOINC porting

Postby John Naylor » Thu Feb 26, 2009 2:37 pm

cenit wrote:don't you think that if PandeGroup had submitted patches to boinc 3 years ago to support GPU, everybody would have benefitted from it? BOINC is an open platform and everybody could help writing code! Doing all in this closed-sourced way has only duplicated efforts!

Google "OpenMM". Any project that wishes to use GPUs (including those that use BOINC, I presume), can use that as a kind of bolt-on, which was developed from knowledge gained while developing the Folding@home GPU and GPU2 clients.

EDIT: I don't think that Beberg believes the F@H network is superior to BOINC, it's just that they work in different ways under the same broad umbrella of Distributed Computing, and the way BOINC works is not necessarily the best way for this project to work. (and I should have mentioned OpenMM is still in development, which reflects the work-in-progress nature of the GPU client's analysis software... and it only supports Molecular analysis... hmmmmm).
User avatar
John Naylor
 
Posts: 950
Joined: Mon Dec 03, 2007 4:36 pm
Location: University of Birmingham, UK

Re: BOINC porting

Postby cenit » Thu Feb 26, 2009 3:11 pm

it's different. OpenMM is for the project, not for the platform!
BOINC has always supported applications that used "internally" the gpu, but now allows also the use of the GPU as a coprocessor if available, transparently for the user, without requiring user interaction! It's different, am I clear enough?

edit: to be more clear: BOINC is a "manager", not a project. It's really simple to port application to it and it would benefits all of us which run ALSO other projects!
User avatar
cenit
 
Posts: 33
Joined: Sat Dec 27, 2008 3:13 pm

Re: BOINC porting

Postby MtM » Thu Feb 26, 2009 3:17 pm

cenit wrote:I've tried the fah client, in fact I give FAH my ATi 4870, at least until a good project under boinc will support it. I've asked this for this reason: could it be that I'm the ONLY ONE which will leave F@H that day? I'm sure I'm not the only one
In fact, I would like to manage all in a single place, it's better if you don't want to support only one project. If I revert the question, have you ever tried boinc? It's really superior to the fah client!
And for the progress on boinc, don't you think that if PandeGroup had submitted patches to boinc 3 years ago to support GPU, everybody would have benefitted from it? BOINC is an open platform and everybody could help writing code! Doing all in this closed-sourced way has only duplicated efforts!
(for the ps3, boinc is only a platform, virtually running everywhere (it runs on the ps3, some projects support it). It stands on the projects to support as many clients as possible)


It's really superior to the fah client! :eo

You're wrong. Boinc is a project manager not a client itself. And the f@h clients are superiour to the boinc clients, not the other way around. How long ago did boinc add cuda support? Does it offer Brook+/stream support? How long is it been offering SMP support? And even then, how many active projects are now using SMP through bionc?

cenit wrote:
toTOW wrote:I think one of the reason that FAH is not available on BOINC has to do with the rapid turnaround that FAH need.
PandeGroup has always said that the faster the WUs are returned, the faster the project makes progress. BOINC software allows you to share your computing power between various projects, which might slow the FAH turnover. BOINC software makes sure that the deadlines are met, but we all know that in FAH, it's better to return the WU as fast as possible ;)


This seems the biggest reason to me, but in a different view: I'm sure that Pande Group are feared to loose clients. I know so many people that don't know anything about distributed computing, but 70% of people who donate do D.C. didn't know the existance of BOINC and the projects that run on it. When they know about it, some leave (yes!) and some runs now more projects, leaving FAH ultimately with less resources. But you cannot say FAH is the best project and so they did something stupid. You can't, in any way!


No it's not the best project, how would one value a project? They all do good things for a good cause.

BUT, they do have the best developed cores/clients right now, see the first part of my post.

F@h loosing donors to boinc will only be misinformed donors who fall for the lure of a nice centralized manager to view progress, I realise most donors don't really look into the backgrounds other then 'what their told and seems to make sence' but those who do, will not move to boinc untill they can match the core's/clients performance.

Don't get me wrong, I don't dislike Boinc. I would like to see something as a boinc manager implemented in f@h ;) An easier setup procedure, a single gui for tracking and managing multiple clients, and even an option to opt which project ranges one would prefer to fold ( so people who fold for huntington for instance, can focus their attention on those if available, and someone more interested in alzheimer would be able to focus their efforts there ). I'd rather have those few advantages boinc offers added to the folding at home program then the other way around.

Btw

BOINC has one very specific model for the types of projects it supports. Folding@home does not fit into that model. That doesn't make it bad, just different and thus incompatible.


does not state it's better, it states it has a diffrent approach. You're trying to make it more it seems but it isn't. There is no one saying this is better or that is better ( besides you claiming boinc was superior in it clients something which I had to correct you on ). Both f@h and boinc are trying to cure deseases, it would be great to have the best of both combined, but it's not fair to make false claims.

Edit: John beat to some of my points, took to long to type and cut up the previous posts ;)
MtM
 
Posts: 3054
Joined: Fri Jun 27, 2008 2:20 pm
Location: The Netherlands

Re: BOINC porting

Postby 7im » Thu Feb 26, 2009 3:25 pm

I compliment you on donating to a worthy cause in FAH. But I don't know how long you have been with fah, and may not know the project as well as some others here.

The different model Beberg referred to above is the time sensitive nature of the fah work units, and the need to submit each work unit as quickly as possible. BOINC on the other hand was designed from the start to allow people to take as along as they wish to complete a work unit, with no indication of urgency, and support for deadlines were only added after fah tried working with the BOINC developers several years ago.

And yes, I have tried BOINC. Is was a piece of junk and the reason I left SETI I to join this project.
User avatar
7im
 
Posts: 14648
Joined: Thu Nov 29, 2007 4:30 pm
Location: Arizona

Re: BOINC porting

Postby cenit » Thu Feb 26, 2009 4:00 pm

7im wrote:I compliment you on donating to a worthy cause in FAH. But I don't know how long you have been with fah, and may not know the project as well as some others here.

Some years now... I've started with an old P3 667 MHz back in the days...

7im wrote:And yes, I have tried BOINC. Is was a piece of junk and the reason I left SETI I to join this project.

Maybe it WAS a piece of junk. But now it could be different.
Please note that I don't stand behind SETI, I don't like their project often confused with the whole boinc.

MtM wrote:It's really superior to the fah client! :eo
You're wrong. Boinc is a project manager not a client itself. And the f@h clients are superiour to the boinc clients, not the other way around. How long ago did boinc add cuda support? Does it offer Brook+/stream support? How long is it been offering SMP support? And even then, how many active projects are now using SMP through bionc?

Sorry if you're so precise, man! I considered boinc as the client and projects as the core... It could stand, not?

MtM wrote:No it's not the best project, how would one value a project? They all do good things for a good cause.
BUT, they do have the best developed cores/clients right now, see the first part of my post.

I think you didn't read my first post.

MtM wrote:Don't get me wrong, I don't dislike Boinc. I would like to see something as a boinc manager implemented in f@h ;) An easier setup procedure, a single gui for tracking and managing multiple clients, and even an option to opt which project ranges one would prefer to fold ( so people who fold for huntington for instance, can focus their attention on those if available, and someone more interested in alzheimer would be able to focus their efforts there ). I'd rather have those few advantages boinc offers added to the folding at home program then the other way around.

BOINC is what you just said. Nothing more, nothing less.
So it seems you don't disagree with me, now...

MtM wrote:Btw
BOINC has one very specific model for the types of projects it supports. Folding@home does not fit into that model. That doesn't make it bad, just different and thus incompatible.


does not state it's better, it states it has a diffrent approach. You're trying to make it more it seems but it isn't. There is no one saying this is better or that is better ( besides you claiming boinc was superior in it clients something which I had to correct you on ). Both f@h and boinc are trying to cure deseases, it would be great to have the best of both combined, but it's not fair to make false claims.


BOINC is not trying to cure disease, now you're wrong!
More important, I donate also to projects that are not involved in medical/cure.
User avatar
cenit
 
Posts: 33
Joined: Sat Dec 27, 2008 3:13 pm

Next

Return to Scientific discussions (Non-FAH)

Who is online

Users browsing this forum: No registered users and 1 guest

cron