BOINC TBD: A New Model - Overclock.net - An Overclocking Community

Forum Jump: 

BOINC TBD: A New Model

Reply
 
Thread Tools
post #1 of 44 (permalink) Old 06-06-2017, 05:43 AM - Thread Starter
New to Overclock.net
 
mmonnin's Avatar
 
Join Date: Nov 2012
Posts: 5,591
Rep: 268 (Unique: 129)
I'm subscribed to one of the BOINC email distributions and there was a link in one about a new model of distributed computing. I didn't see a reference to this anywhere else so I'm starting one here.

http://boinc.berkeley.edu/tbd.php

It explains some issues with the current BOINC model:
Quote:
The model didn't work as envisioned, for a number of reasons:
  1. Creating and operating a VC project is harder than we realized: it requires a combination of resource and skills (Win/Mac programming, sysadmin, DB admin, web design, PR/outreach) that few academic research groups have.
  2. For a research group, trying to use VC is a risk. There's a substantial investment, with no guarantee of any return, since no one may volunteer. Adding a VC component to a grant proposal adds uncertainty and weakens the proposal.
  3. The computing needs of many research groups are sporadic - e.g. they need a big chunk of throughput every now and then. For such groups, buying computing time on a commercial cloud may be cheaper than using VC.
  4. Attracting volunteers is a marketing exercise. It's difficult to do effective marketing when there are dozens of competing brands (i.e. projects names).
  5. Most volunteers aren't interested willing to survey and assess a large set of projects once, much less repeatedly.
  6. We made little effort to interface, technically or politically, with the mainstream HPC/HTC world (Grid, Supercomputing, Condor, etc.). They came to view VC in negative ways: as a threat, a gimmick, etc. Around 2006 there was a brief and small interest in VC in academic computer world. Since then, nothing: no conferences on distributed computing list VC as a topic of interest. This has been damaging to VC; e.g. no one is working on solving the hard problems that arise in VC (such as how to grant credit).

A New Hope Model:
Quote:
  1. Partner with existing HTC computing providers such as supercomputing centers and science portals to add BOINC-based back ends. These projects would be operated by the provider's staff. Tens of thousands of scientists use such computing providers. These scientists would benefit from lower queueing delays, higher throughput and lower cost. But they wouldn't need to do anything; they wouldn't even need to know that VC is being used.
  2. Create an account manager (let's called it "TBD" for now) acting as the primary volunteer interface. TBD lets volunteers express their preferences in terms of keywords (scientific areas and locations) rather than selecting specific projects. Based on these preferences, and corresponding keywords of projects and applications, TBD dynamically assigns computers to a set of vetted projects, which would include both existing (single-group) projects, as well as the new computing-center projects.

Quote:
  1. It doesn't interfere with or preclude existing BOINC activities. Current projects continue as they are. Scientists can create new single-group projects if they want. Volunteers can attach individual projects as they currently do, or use existing account managers like BAM! and Gridrepublic.
  2. TBD will act as an allocator of computing power. This will be based in part on user preferences, but there will of necessity also be a higher-level allocation policy, decided on by an organization. The decision process should include merit and need; it may include politics and money as well. NSF has an organization - XSEDE - that does this for NSF-funded computing resources. I'm in contact with XSEDE, and hope to include them in TBD. Involving NSF in the process is important; but this project needs to be international. This part of the model needs to be worked out at a high level.
  3. The model focuses on large HPC-provider-level projects, but it actually encourages single-group projects, since they can apply for an allocation from TBD and be assured of computing power prior to making any investment.
  4. TBD can serve as a brand for VC marketing purposes. It will also provide a basis for corporate partnerships; if technology or game companies want to support VC, they can support TBD rather then having to select individual projects.

My comments:
I don't think there will be extra millions of people that will run this model. DC/VC/FAH/BOINC costs electricity to run and at 100% processor utilization many Dell/HPs/etc fans start to get loud. With those two things you're just be left with the dedicated who are already running BOINC as is. Probably some more but I'm not sure millions more.

'For the Science' is nice and all but let's be honest. It's a competition. There are points involved. Some homogeneous blend of points across the group of projects won't be fun. And it won't be easy to get work from one project to give out the same points as another. Some projects get more or less users due to their high or low PPD output as is. That's a clear indication of the value of points and competition.

Control. I'm guessing we'd lose control over exactly what we run and are left to groups like math, astronomy, medical. Some projects stress hardware much more than others. We'd have to tune for the hardest and lose efficiency on another. Then there's the memory/disk requirements as well. How does one configure it like we can BOINC now.

I do see the benefit of a generic computing grid where smaller projects with intermittent work can get the computing cycles they need.


mmonnin is offline  
Sponsored Links
Advertisement
 
post #2 of 44 (permalink) Old 06-06-2017, 06:38 AM
New to Overclock.net
 
BeerCan's Avatar
 
Join Date: Sep 2012
Location: Florida USA
Posts: 630
Rep: 25 (Unique: 14)
If I am comprehending that correctly I am out. I also think many others would leave as well.



BeerCan is offline  
post #3 of 44 (permalink) Old 06-06-2017, 07:00 AM
New to Overclock.net
 
Egilman's Avatar
 
Join Date: Dec 2015
Location: Tuvalu
Posts: 1,114
Rep: 34 (Unique: 18)
Quote:
Originally Posted by mmonnin View Post

I'm subscribed to one of the BOINC email distributions and there was a link in one about a new model of distributed computing. I didn't see a reference to this anywhere else so I'm starting one here.

http://boinc.berkeley.edu/tbd.php

It explains some issues with the current BOINC model:
A New Hope Model:

My comments:
I don't think there will be extra millions of people that will run this model. DC/VC/FAH/BOINC costs electricity to run and at 100% processor utilization many Dell/HPs/etc fans start to get loud. With those two things you're just be left with the dedicated who are already running BOINC as is. Probably some more but I'm not sure millions more.

'For the Science' is nice and all but let's be honest. It's a competition. There are points involved. Some homogeneous blend of points across the group of projects won't be fun. And it won't be easy to get work from one project to give out the same points as another. Some projects get more or less users due to their high or low PPD output as is. That's a clear indication of the value of points and competition.

Control. I'm guessing we'd lose control over exactly what we run and are left to groups like math, astronomy, medical. Some projects stress hardware much more than others. We'd have to tune for the hardest and lose efficiency on another. Then there's the memory/disk requirements as well. How does one configure it like we can BOINC now.

I do see the benefit of a generic computing grid where smaller projects with intermittent work can get the computing cycles they need.

Essentially as I read that, they are saying that they cannot trust the end user to determine what he/she wants to run. So in a virtual way, they will decide for us what should be run on our hardware. For example, Acoustics is a program that is trying to sharpen sound detection under the sea. to allow for better mapping of objects and textures under water. Now if the project was devoted to the improved detecting of targets underwater, very few in the DC crowd would run it as that is a military application. Since they (the project) says it is not a military application many people run it. despite the clear and obvious application to military purposes. That is called trust.

In a virtual project, you don't even get to ask the question......

Quote:
Originally Posted by BeerCan View Post

If I am comprehending that correctly I am out. I also think many others would leave as well.

I'm with you.....


Egilman is offline  
Sponsored Links
Advertisement
 
post #4 of 44 (permalink) Old 06-06-2017, 07:05 AM - Thread Starter
New to Overclock.net
 
mmonnin's Avatar
 
Join Date: Nov 2012
Posts: 5,591
Rep: 268 (Unique: 129)
It mentions leaving the current BOINC projects as is and they would still exist. It seems to be better suited towards one off computer models that need compute power for a short period. Those projects that currently pay for TACC services. Maybe for the layman computer user than those of us here that build our own computers.


mmonnin is offline  
post #5 of 44 (permalink) Old 06-06-2017, 07:05 AM
New to Overclock.net
 
BeerCan's Avatar
 
Join Date: Sep 2012
Location: Florida USA
Posts: 630
Rep: 25 (Unique: 14)
One of the reasons I dislike [email protected] is because I never know what I am crunching. This seems like they want to move in that direction, boinc just crunches wu's and you never really know what the science is. And as EG say's it will also remove a layer of trust.



BeerCan is offline  
post #6 of 44 (permalink) Old 06-06-2017, 07:08 AM - Thread Starter
New to Overclock.net
 
mmonnin's Avatar
 
Join Date: Nov 2012
Posts: 5,591
Rep: 268 (Unique: 129)
Quote:
Originally Posted by BeerCan View Post

One of the reasons I dislike [email protected] is because I never know what I am crunching. This seems like they want to move in that direction, boinc just crunches wu's and you never really know what the science is. And as EG say's it will also remove a layer of trust.

I don't think there is any more distance between FAH work and say [email protected] work. [email protected] says I'm crunching some data from a Fermi telescope. FAH I can actually see the protein. This is even a further step back. I give a computer access and one day it may be working on FAH and tomorrow [email protected]


mmonnin is offline  
post #7 of 44 (permalink) Old 06-06-2017, 07:10 AM
New to Overclock.net
 
BeerCan's Avatar
 
Join Date: Sep 2012
Location: Florida USA
Posts: 630
Rep: 25 (Unique: 14)
Quote:
Originally Posted by mmonnin View Post

I don't think there is any more distance between FAH work and say [email protected] work. [email protected] says I'm crunching some data from a Fermi telescope. FAH I can actually see the protein. This is even a further step back. I give a computer access and one day it may be working on FAH and tomorrow [email protected]

Yeah, I guess your right. I just feel a little more connected in boinc because of the way you can pick projects and move around. With [email protected] it's pretty much set and ignore/forget.



BeerCan is offline  
post #8 of 44 (permalink) Old 06-06-2017, 07:18 AM - Thread Starter
New to Overclock.net
 
mmonnin's Avatar
 
Join Date: Nov 2012
Posts: 5,591
Rep: 268 (Unique: 129)
Quote:
Originally Posted by BeerCan View Post

Yeah, I guess your right. I just feel a little more connected in boinc because of the way you can pick projects and move around. With [email protected] it's pretty much set and ignore/forget.

The best example FAH has is proteins related to Huntington, Alzheimer's, etc and supposedly being able to chose. I guess that could be the equivalent of different apps w/in a single BOINC project.


mmonnin is offline  
post #9 of 44 (permalink) Old 06-06-2017, 07:42 AM
New to Overclock.net
 
Egilman's Avatar
 
Join Date: Dec 2015
Location: Tuvalu
Posts: 1,114
Rep: 34 (Unique: 18)
Quote:
Originally Posted by mmonnin View Post

The best example FAH has is proteins related to Huntington, Alzheimer's, etc and supposedly being able to chose. I guess that could be the equivalent of different apps w/in a single BOINC project.

Well, it is just another indication of the direction DC is taking. one in which I'm not at all happy with.

Two things bother me...

The monetization of Boinc, (reference the gridcoin project)

The removal of choice, (reference the virtualization of boinc)

DC is a giant supercomputer, one that the sciences and industries have no control over, much like say the Aracaibo telescope not having a reservation system for it's time, just anyone can come up and point it at whatever you want. Doesn't work very well in real life for doing actual science.

Well DC right now is much the same way, people can run what they want when they want to, which is the way I like it. Take that away and you become just another cog in the machinery.......

And I do not envision myself as just another gear tooth.

Science is becoming altogether much too socialist for my tastes.... (and gridcoins bid to take over boinc with the acquiescence of Berkeley is a whole 'nother conversation)

No, I don't like the overall direction DC is taking.....


Egilman is offline  
post #10 of 44 (permalink) Old 06-06-2017, 07:49 AM - Thread Starter
New to Overclock.net
 
mmonnin's Avatar
 
Join Date: Nov 2012
Posts: 5,591
Rep: 268 (Unique: 129)
I dislike gridcoin more than this trend. Still not sure where the money comes from when crunching for projects that don't give in the 1st place. From what I heard there's still someone in between making money off what you crunch. I wish they would split those type of teams from the regular teams on Stats pages like google was on EOC's FAH page.


mmonnin is offline  
Reply

Quick Reply
Message:
Options

Register Now

In order to be able to post messages on the Overclock.net - An Overclocking Community forums, you must first register.
Please enter your desired user name, your email address and other required details in the form below.
User Name:
If you do not want to register, fill this field only and the name will be used as user name for your post.
Password
Please enter a password for your user account. Note that passwords are case-sensitive.
Password:
Confirm Password:
Email Address
Please enter a valid email address for yourself.
Email Address:

Log-in



Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)
 
Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may post new threads
You may post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off