PDA

View Full Version : Suggest a distributed computing project



ShinAce
2012-Jun-15, 04:17 PM
A long time ago, I participated in Seti@home. Well, they eventually got to the point where every piece of work was redone by over 3 people so the results became redundant. I then moved on to Folding@home, but gave that up as part of a reformat.

Last night, I reinstalled the BOINC program and was going to try MilkyWay@home. I will not be continuing work for MilkyWay@home due to it being a waste of my CPU time. It downloaded 6 units of work, completed them overnight, then would not fetch new work or return results. I decided to run the benchmark in BOINC and suddenly it restarted all the units from 0%. So, after 12 hours of CPU time I have contributed nothing to MilkyWay@home. It will stay that way since I am dissapointed in their work server.

Does anyone have any projects that they think are interesting but also, more importantly, a project which is actually doing work? None of this university hosted servers garbage. I want a full fledged project that's more than a few people kicking a can around.

Ara Pacis
2012-Jun-16, 08:47 AM
Are they still doing folding@home? It was for protein folding calculations.

headrush
2012-Jun-16, 12:13 PM
You could try out LHC@home (http://lhcathome.web.cern.ch/LHCathome/Physics/)


LHC@home 2.0 is an extension of the LHC@home platform, launched in 2004 to help physicists simulate protons beam dynamics in the LHC. At that time, doing full-fledged simulations of particle collisions was beyond the scope of volunteer computing. But the evolution of computer software and hardware, and in particular the use of virtual machine technology, has enabled a breakthrough...LHC@home 2.0!

It involves setting up a virtual machine though.

ShinAce
2012-Jun-17, 03:28 PM
Folding@home must still be going. I've donated a year of crunch time already. I love the idea, but their work is like children playing with lego. Such simple proteins. Due to that, they'll probably study mostly viruses.

I didn't know about LHC2.0! Thanks. I had also given time to LHC1.0 until the lhc started up.

headrush
2012-Jun-17, 07:27 PM
I didn't know about LHC2.0! Thanks. I had also given time to LHC1.0 until the lhc started up.

I'd be interested in how you get on with it if you go that way. I have started installing it, but ran into issues due to 64 bit / 32 bit conflicts on my machine. The LHC vm is 32 bit which should be fine but the boinc client also has to be 32 bit which is an issue.
I'm running Fedora 15 x86_64. It's a pity because I have 8 cores going mostly idle.

I don't think it would be very efficient to virtualise the whole thing, but it may work better that way, (64bit machine > 32 bit guest > 32 bit boinc > 32 bit vm).

TrAI
2012-Jun-18, 03:06 AM
A long time ago, I participated in Seti@home. Well, they eventually got to the point where every piece of work was redone by over 3 people so the results became redundant.

Well, you will probably have a hard time finding any distributed computing project that will not submit the same work to three or more clients, as this is pretty much standard practice in distributed computing these days. The reason is pretty simple; you want to ensure the predictability of project progress and the integrity of the results before adding it to your database of final results.

This sort of redundancy is quite important, as any number of problems may occur, the data may have been corrupted in some way, the client might have some issue or buggy hardware or software that makes the results inaccurate, the client user might hack the client to falsify data in an attempt to get more credit(SETI@Home apparently did have problems with this before, and it is partly responsible for the development of the BOINC system and how it is implemented with regards to redundancy and security), and, of course, the client might go off-line for a while, and so your project would loose time by having to resubmit the work units.


I then moved on to Folding@home, but gave that up as part of a reformat.

You might be able to continue the work on this project, though it does not use BOINC now, it is still running with its own client.


Last night, I reinstalled the BOINC program and was going to try MilkyWay@home. I will not be continuing work for MilkyWay@home due to it being a waste of my CPU time. It downloaded 6 units of work, completed them overnight, then would not fetch new work or return results. I decided to run the benchmark in BOINC and suddenly it restarted all the units from 0%. So, after 12 hours of CPU time I have contributed nothing to MilkyWay@home. It will stay that way since I am dissapointed in their work server.

You are a bit impatient, aren't you? 12 hours is not enough to give a good picture of how reliable the project is, anything may happen in such short amounts of time.

I would suggest trying out a project for a month or more before abandoning it on these grounds. You will be unlikely to loose any progress either, as your client will generally automatically submit the results and fetch new work next time it can establish contact with the project servers. I subscribe to several projects(SETI, Einstein, Rosetta and World Community Grid) and so even if one were to be off-line, there would still be something for my machine to do, generally it is running about two work units from each project at any given time.

It's sad that you lost the progress with your current work units, but it was only a few hours worth of work, so it wouldn't have taken that long to redo them. I think it is best to have a laid back approach to these projects, it is not really something you do for gain. Sure you get points, and maybe, if your client gets a lucky work unit, you get some mention or something, but there will always be some users with a farm of computers(or a whole company/campus worth of computers under their power), who can crunch through more work in a day than the regular home user can in a year, so the points do not really matter much. It is more about donating computing power to help some project progress because you have some interest or belief in the importance of the work.


Does anyone have any projects that they think are interesting but also, more importantly, a project which is actually doing work? None of this university hosted servers garbage. I want a full fledged project that's more than a few people kicking a can around.

Well, most of the larger distributed projects are run by universities and research institutes, but some projects are run by other organizations. You could look over the Wikipedia list of distributed computing projects (http://en.wikipedia.org/wiki/List_of_distributed_computing_projects) to see if there is something that looks interesting. The World Community Grid project is the only one of the projects I subscribe to that is not directly run by a university(or similar), as it is operated by IBM. It hosts many sub-projects related to drug/disease research and some environmental research, of course, these projects are generally hosted by IBM for universities.

The largest projects are likely to be fairly stable, but still, there is always some possibility of problems, so my suggestion is to find a couple or more projects to subscribe to, then you will generally always have something running.

Ara Pacis
2012-Jun-18, 07:57 AM
Are they still doing folding@home? It was for protein folding calculations.D'oh! I just realized that you wrote that in the OP. I thought I double-checked. Odd, that.

ShinAce
2012-Jun-18, 07:05 PM
I'd be interested in how you get on with it if you go that way. I have started installing it, but ran into issues due to 64 bit / 32 bit conflicts on my machine. The LHC vm is 32 bit which should be fine but the boinc client also has to be 32 bit which is an issue.
I'm running Fedora 15 x86_64. It's a pity because I have 8 cores going mostly idle.

I don't think it would be very efficient to virtualise the whole thing, but it may work better that way, (64bit machine > 32 bit guest > 32 bit boinc > 32 bit vm).

Hi Alan. My name is also Alain(literally the same name but in french).

I downloaded virtualbox thinking it would be no big deal(windows vista), but had issues like you did. Install was fine, creation of a virtual machine went without errors, but it's not liking the virtual hard drive. Back in the day, I would have given it carte blanche on the spare partition, but I only have one disk drive on a single partition. Didn't feel like spending who knows how many hours to get it going. It's too bad, because I love the work the LHC is doing. I'm actually studying physics myself.

As a screensaver, it's a million times more attractive. Just install a small program, let it use my idle time, and I can move on to real things. Isn't that what made it so popular in the first place? How many people want to spend this kind of effort in order to stress their CPUs and motherboards? No thanks.

I've decided I want to go back to screensaver mode. I'll be running Einstein@home, looking for pulsars

headrush
2012-Jun-21, 04:45 PM
Hmmm, that's interesting. I'll have another bash at it at the weekend and see if I can get it running at all.

HenrikOlsen
2012-Jun-22, 11:46 AM
Folding@home must still be going. I've donated a year of crunch time already. I love the idea, but their work is like children playing with lego. Such simple proteins. Due to that, they'll probably study mostly viruses.
Their work is simple because computational chemistry is still in its infancy, they need to work on simple proteins to get the theory right first.

They aren't children playing with Lego, they haven't reached that level yet.
They're infants playing with over-sized cubes.

Viruses are probably way too complex yet.

HenrikOlsen
2012-Jun-22, 11:49 AM
Well, you will probably have a hard time finding any distributed computing project that will not submit the same work to three or more clients, as this is pretty much standard practice in distributed computing these days. The reason is pretty simple; you want to ensure the predictability of project progress and the integrity of the results before adding it to your database of final results.
Actually it's even simpler, people use hacked clients which returns garbage results quickly in order to artificially boost their standing.
It's seen by some as a competition so you will get cheaters.

headrush
2012-Jun-22, 11:10 PM
Actually it's even simpler, people use hacked clients which returns garbage results quickly in order to artificially boost their standing.
It's seen by some as a competition so you will get cheaters.
Good point, which leads to the question of whether the scientists running these things should offer the option of being part of a team (most do in my experience).

TrAI
2012-Jun-23, 03:53 AM
Actually it's even simpler, people use hacked clients which returns garbage results quickly in order to artificially boost their standing.
It's seen by some as a competition so you will get cheaters.

Hmmm, hacked clients was one of the examples I mentioned in the next paragraph of my post, and SETI@home's experience with such problems was important in the implementation of the redundant calculation approach used in the BOINC system.

However, My point was that whether the problem is a hacked client, a buggy math library or processor that hasn't yet been patched, an error in the port of the client to a specific platform or what ever, you want to detect erroneous results in a timely and consistent manner before the data is committed, and so there are several compelling reasons for a distributed computing network having this sort of redundancy.

ShinAce
2012-Jun-24, 02:54 PM
Their work is simple because computational chemistry is still in its infancy, they need to work on simple proteins to get the theory right first.

They aren't children playing with Lego, they haven't reached that level yet.
They're infants playing with over-sized cubes.

Viruses are probably way too complex yet.

LOL! My sentiments exactly. With the villin headpiece simulation complete, they're finally moving on to disease related folding projects. I do believe the power of this technique will be incredible within 25 years. It's just going to be a question of having enough power to do it. Kind of like how I was playing super mario bros in 1986 without even thinking of how realistic games would have become in 2011.

The problem I really have with SETI is that in the beginning, they wondered if they would get enough CPU power. They did, so they broadened the search. Even with that, they had so much CPU time available and such a solid server, that they got to the point where they would rerun old units. These units had already been sent to 3-4 people, were validated, and then resent to keep CPUs busy. Meanwhile, still very few promising candidates for signals. As of today, I know of no signals found by Seti@home. The "WOW" signal sparks imagination, but seti@home squashes it with a solid null result.

LIGO seems like it still has a chance to prove the idea right. Plus the new interferometer is being built. New data to come!

LHC@home is cool but hard to install and difficult for the layman to understand. You're calibrating that big machine thing there for muons? Cool, what's a muon?

World processing power is only going to increase. Moore's law isn't taught because it's garbage. It's a good rule. Soon enough, a decent world climate simulation won't need a cluster to compute in time.