NSF funds new model for BOINC

Message boards : News : NSF funds new model for BOINC

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4

AuthorMessage
boboviz

Send message
Joined: 12 Feb 11
Posts: 226
Italy
Message 79102 - Posted: 19 Jun 2017, 10:18:00 UTC

A little doubt. In which way the work queue are managed?
I mean: for example, there will be the "medical" section of TBD with 5 project/universities/research centers that partecipate. Which will have the priority of work? The others have to wait the finish of work of one project to "access" to our (volunteers) computers???
I predict furious quarrels :-P
ID: 79102 · Report as offensive    Reply Quote
vlado101

Send message
Joined: 29 Jan 15
Posts: 14
United States
Message 79123 - Posted: 19 Jun 2017, 20:50:02 UTC - in response to Message 79102.  

A little doubt. In which way the work queue are managed?
I mean: for example, there will be the "medical" section of TBD with 5 project/universities/research centers that partecipate. Which will have the priority of work? The others have to wait the finish of work of one project to "access" to our (volunteers) computers???
I predict furious quarrels :-P


Hi boboviz,

That maybe be true that proficient BOINC users such as yourself who join and research projects that they want to contribute might see this as a problem, but if you have someone who has not heard of BOINC before and has no idea of what projects are available (much less the scientific benefit) why would they care? To them they are contributing the medical research. I think the whole point of the new model is to get people who have never heard or thought about contributing their computing time to join. The sad fact of that is that a program would need to tailor to the lowest denominator, aka people who have not heard of BOINC or have done prior research on which projects exist. Would it not be better that those projects receive at least a little bit of this un-used computing power than none at all? The argument about how the projects would be weighted is beside the point for this type of user. They would need to be engaged, I am sure either access to papers or some new form of badges, but I see that the majority of people that this model is trying to attract would not do the same form of research. However, from my understanding the other model of joining projects (aka you choosing projects will exist). If the hypothetical initial user would then wonder which projects and how much of it is weighted he or she would then be able to join other projects that he or she would feel more of a connection to.

Please correct me if I am wrong, however I do feel that there is room for the new approach and the current BOINC model. I say all the more power to have people join and donate their computing time and it would be worth a try.
ID: 79123 · Report as offensive    Reply Quote
boboviz

Send message
Joined: 12 Feb 11
Posts: 226
Italy
Message 79124 - Posted: 19 Jun 2017, 21:58:04 UTC - in response to Message 79123.  

The argument about how the projects would be weighted is beside the point for this type of user. They would need to be engaged, I am sure either access to papers or some new form of badges, but I see that the majority of people that this model is trying to attract would not do the same form of research.


My post is to move the focus from volunteers to projects admins.
Why a project have to partecipate to TBD and not to Boinc, for example? Advantages?
How TBD administrates the queues of work? It's not only a "front-end" problem/volunteers, but also back-end
ID: 79124 · Report as offensive    Reply Quote
Profile Carl Christensen

Send message
Joined: 20 Oct 06
Posts: 33
United States
Message 79145 - Posted: 20 Jun 2017, 21:09:08 UTC - in response to Message 78746.  
Last modified: 20 Jun 2017, 21:13:02 UTC

>>It's your fault. With good and continously updated documentation, support on forum/mailing list/etc, the administration of a VC >>project for a research group is NOT so hard.

That's total bullshit -- having run two fairly big BOINC projects (CPDN & QCN) and helped with many others - I think I can state unequivocally that 99.9% of research groups don't have the staff to properly start and support a BOINC project. The only ones that succeeded were ones that were comfortable looking outside the science post-doc staff to actually get real (and good) computing people. And extra credit if they actually had scientists that new anything about computers and software (e.g. Bruce Allen).

Also you need a project leader with a good bit of "chutzpah" and zeal to get publicity etc. That narrows down the potential field a lot. There should probably be a "Drake Equation" for BOINC --- yeah the potential pool of scientists making projects is a million but when you get down to it you end up with a handful.

So it seems to me that BOINC has well saturated the current "market" or model and the 'good ole projects' will be unaffected so there needs to be something new like this to expand things (or just keep things alive).
ID: 79145 · Report as offensive    Reply Quote
Profile David Anderson
Volunteer moderator
Project administrator
Project developer
Avatar

Send message
Joined: 10 Sep 05
Posts: 621
Message 79222 - Posted: 23 Jun 2017, 4:48:45 UTC
Last modified: 23 Jun 2017, 4:49:30 UTC

My aggregate responses in 4 areas: TBD, BOINC, cryptocurrencies, and myself.

TDB:
- TBD is not a BOINC replacement. It's a new project manager for volunteers who
want to use it, and it's technology to help HPC centers add BOINC back ends.
- The role of TBD is not to maintain BOINC or GPUGrid; I'm not sure where that confusion arose.
- TBD doesn't handle jobs. Please read the proposal and about account managers.
- Supercomputer centers are not commercial.
They're government-funded, and scientists apply for resource allocations.
TACC and SDSC (San Diego Supercomputer Center) are examples.
The function of TBD (and volunteer computing in general) is to save the government money
(important, as funding agency budgets are being cut) and to allow novel research to be done
that otherwise couldn't afford the needed computing.
- In some areas, like biomedicine, research results may eventually get commercialized (e.g. drug discovery).
That's the nature of academic research.
People need to weigh this against the benefit of the outcome (e.g. having a drug available).
- TBD will tell volunteers where each job came from, if they want to know.
It won't, however, force volunteers to learn about and select projects.
- TBD will vet the projects it encompasses. There will be no "rogue" projects.

BOINC:
- The API is, of course, under version control. If there are bugs, file Github issues.
- BOINC is scarcely "hindering" GPUGrid. We've provided every feature they've requested,
and they haven't requested any for at least 5 years.

Cryptocurrencies:
- I'd love to see BOINC credit used as proof of work for a cryptocurrency.
In fact I proposed this to Bitcoin a long time ago;
they declined because BOINC isn't fully distributed; fair enough.
But this can't be built directly into BOINC; NSF wouldn't be OK with that.

Myself:
- I was interested to read that I moved to the private sector during the gap in BOINC funding.
This is an example of "fake news", and it discredits the originator.
Actually, I've been working mostly on SETI@home,
developing a new back-end system called Nebula
that completes the ET-finding pipeline,
also a bit on HERA, a new radio telescope that will image the extremely early universe.
Neither of these is private sector, of course.

Did I leave out anything?
-- David
ID: 79222 · Report as offensive    Reply Quote
Profile pschoefer

Send message
Joined: 5 Aug 06
Posts: 47
Germany
Message 79280 - Posted: 25 Jun 2017, 9:23:29 UTC - in response to Message 79222.  

Thanks for the clarifications, the idea looks much better to me now.

Another point that crossed my mind is that those volunteers who are attracted by the "TBD" approach will likely also be interested in seeing actual results of the research they support:
While I understand that not every single task of every project is meaningful considered by itself, but the project may need many years of computation until a paper can be written, published results often slipped under the radar in the past and volunteers complained that all their effort would lead to nothing. I can imagine that the new HPC center projects will contribute to papers quite regularly, because they can also host short-term projects, and I think that they should definitely communicate their successes to the volunteers (a news item propagated to the BOINC Manager might not be enough; I myself would be happy with an email notification, GPUGRID's approach of awarding badges based on the contribution to particular papers might also be quite attractive). And, of course, that information would need to be propagated to "TBD" if you don't want that the volunteers have to deal with the individual projects. Having a list of papers the volunteer has made possible with his resources would likely be much more popular than just being able to say "hey, I'm contributing to biomed research".

And just a few words on cryptocurrencies: While I agree that it is a shame to see all that wasted computing power, I predict (based on psychology and a bit of history) that it would be trading short-term benefit for another can of worms and the beginning of the end of BOINC as a Volunteer Computing platform if it was built into BOINC, so it's good that this is not going to happen.
ID: 79280 · Report as offensive    Reply Quote
Previous · 1 · 2 · 3 · 4

Message boards : News : NSF funds new model for BOINC


Copyright © 2017 University of California. Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.