Distributed Computing "Advances"

Quirk writes "NewScientist is reporting on..."Software to be launched in January will let PC users run as many "distributed computing" projects as they like. The program will let PC users search for aliens, help predict climate change and perform advanced biological research - all at the same time."'It is called the Berkeley Open Infrastructure for Network Computing (BOINC). BOINC acts like a software platform that can run a number of screen-saver style applications on top of the PC's own operating system.'"

Who is Benefiting?

[Famatra]

I was interested in the folding protein project, but are the results open to the public (like the human geneome project) free of charge, or will someone making a buck off *my* computing power?

With all the distributed computing projects out there be sure to read the fine print, if your going to use your computer for a project make sure its helping everyone instead of a few corporations make $.

Re:Wont we get this in longhorn with...

[Moderation abuser]

NUMA is great for dedicated machines, but general purpose machines lending out RAM to other systems? Get real, you'd be better off with a BFO page space.

Remote RAM has to be instantly available and it can't go away. Shitty isn't the word for it when we're talking about using general purpose networking kit like gigabit for NUMA. Utterly unusable and waste of time are the best words to describe it. You need SCI, Myrinet or similar to get shitty performance.

Multiple Projects on the same machine

[Seek_1]

This really isn't as good as you might think.

Most distributed computing projects are distributed because they need massive amounts of CPU cycles. Running multiple projects on one machine isn't going to make the projects faster since the same amount of CPU cycles are now being divided up amongst the number of projects that you're running. Infact it'll actually be less because now the machine has to deal with the overhead of switching between project processes.

On the other hand it might make sense if you were running a CPU-intensive project and a data-intensive project at the same time (ie projects that will maximize separate non-conflicting resources on the same machine..)

My Folding@Home Team [stanford.edu]

Distributed Computing OR my time is NOT free

[Anonymous Coward]

Well....the processors in my computers are OWNED by me. I pay the electricity bills to operate them, and YOU want to use my processor time for FREE ?? I dont think so, pony up some cash or keep your distributed clients, thank you.

Re:Double work

[jkcity]

Double results and checking also helps to capture random errors i would guess as well though, not just cheating.

Flexibility at the cost of speed?

[Dan Crash]

I know one of the reasons they created BOINC is that the current SETI@home clientbase is very rigid and can only process data from one telescope -- Aricebo. I also know that the commandline client is tons faster than the screensaver-based client. Is BOINC's flexiblity going to end up making BOINC clients slower than the current dedicated clients?

Re:Distributed Computing OR my time is NOT free

[Seek_1]

I suppose contributing to the collective good doesn't turn your crank then does it?

Re:Flexibility at the cost of speed?

[Loosewire]

is that that the screensaver is only less efficient when displaying its graphs?? how does it compare to the command line versin once the screensaver switches to go to blank mode (thats a setting directly in the screensaver not just a power save of the monitor....)

Java Applet distributed computing

[DeepDarkSky]

I thought a long time ago, why not make distributed computing applications as Java Applets hosted on web servers?

Pros:
- Nothing to "install".
- Cross platform (write it once, run it everywhere, right?)
- Easy to use (just browse)

Cons:
- Speed.
- Full featured screen saver not possible?
- uh...speed?

Re:Skeptical

[RicktheBrick]

This only works if one is using one's computer for personal reasons and contributing the wasted cyles to the cause. If one leaves the computer on only to do the calucations than one is paying more for the electricity to run the program than the calculations are worth. I refer to the $5 million dollar supercomputer at Virginia Tech. This computer can do 8 trillion calculations a second. Now how many pc would it take to equal that and than caculate the cost of electricity for that amount of pc's. Now the question is "Is it better to contribute money to a cause so they can buy and maintain the supercomputers or to run one own computer?" Today it might be questionable but in a couple of years the supercomputers will difinitely be cheaper. Running a supercomputer would eliminate the redundancy problem which is 5 times at Gric.org(cancer and smallpox research).

Re:Multiple Projects on the same machine

[SETIGuy]

Running multiple projects on one machine isn't going to make the projects faster since the same amount of CPU cycles are now being divided up amongst the number of projects that you're running.

The default BOINC operating mode on single processor machines is to alternate projects to balance work between projects.

But that's not really the point. I'll assume you donate to charities. Do you only donate to one charity? Probably not, becase there is more than one worthy charity.

I think that there is more than one worthy distributed computing project as well. One of the design goals of BOINC is to allow volunteers/donors to spread their contribution among worthy projects.

Another goal is to unify the donor base. Projects may have a varying processing load depending upon data source. Some projects may be I/O intensive rather than processor intensive. Some may be network intensive. (Running an I/O intensive and a processor intensive process simultaneously DOES make better use of the machine.) It's no big secret that SETI@home has somewhat more processing capacity than it needs right now. (That may change soon, but that's another story.) BOINC allows projects and their donors to shift resources to where they are needed. Assuming everyone signs up for more than one project, excess processing capacity will flow to where it's needed.