folding@home and Gentoo Linux Users Everywhere

[digicosm]

Hi gang,

Taking a cue from kanuslupus' idea, I formed a team for Gentoo users participating in the folding@home distributed project as well! Unlike SETI, I did not find any Gentoo-specific teams for folding@home, so I created one. I used the same name (Gentoo Linux Users Everywhere, or GLUE -- did you know you created a cool acronym, kanuslupus?!)

Relevant links:

• Folding@home: http://folding.stanford.edu
  GLUE team: http://folding.stanford.edu/cgi-bin/teampage?q=11298

To join this team, simply put '11298' as your team number in the folding client.

Not trying to steal any thunder from the SETI project, but if anyone is folding proteins, it would be fun to aggregate results this way as well.

By the way, it's likely this project will get a huge boost, as Google Toolbar will soon be rolling out a button that will distribute idle cycles to this project. Granted, Google Toolbar is an IE-only thing for now, but it has the potential of adding a lot of users to distributed computing projects.

[pointwood]

Another project that might interest, is Distributed Folding: http://distributedfolding.org/

It's not the same as Folding@Home (F@H) and it is a relatively new project, but it is working on something in the same area as F@H.

I will not create a Gentoo team for it, since I'm already affiliated with the team (Ars Technica Team Stir Fry), but it's not that difficult to create a team :p I would love to see a Gentoo team in the stats

You can find a description of the project on our website, right here: http://tsf.dbestern.net/background.php

I did F@H before I started on this project, I prefer this project since the primary person behind the project is great guy and quite familiar with Linux - the client was originally written on Linux, not Windows. There are clients for a lot of platforms and more will be added soon. Furthermore, the client is a bit faster on Linux than on Windows and the bugs that have been found, has usually been fixed within a day
When I did F@H, the linux client was much slower (about 30%, but it should be much faster now) and the Linux version was named "clientname.exe" Don't know if that's still true.

If you decide to run this client, be sure to read the readme1st.txt that comes with the client. It contains a lot of *very* important info in regards configuration of the client (example: if you got enough RAM, you can add a switch which gives you a 100% performance boost on Athlon PC's and 50% on P4 PC's). You can also check the official forum where the project management frequently visits and posts to.

One warning though, this project is not the best for modem users, since you have to upload relatively much. That will change, either when the next generation of the client is released (not this year) or if he re-writes the current client, then the uploads will be much smaller when the next version arrives (in about a month I would guess).

[pjp]

I used the same name (Gentoo Linux Users Everywhere, or GLUE -- did you know you created a cool acronym, kanuslupus?!)

I did try to make a pronounceable acronym. I didn't put much into it, but GLUE was all I came up with.

[digicosm]

Another project that might interest, is Distributed Folding: http://distributedfolding.org/

It's not the same as Folding@Home (F@H) and it is a relatively new project, but it is working on something in the same area as F@H.

Thanks for the link! I took a look at the site and ran the client for a bit. At first I was amazed at the speed of it, but then I realized by studying the site that it is simply generating random folds by stringing the atoms together in random directions. Nothing is actually being "folded". These billions of different folds are then sent to the server.

I guess I don't see the point, since 99.9999999+% of the actual folds generated will turn out to be bogus and unviable. If a real fold happens to be generated, what exactly does that prove? Sure, one out of a million monkeys hammering on a typewriter will emit Shakespeare, but that doesn't mean that monkey is a genius.

F@H may be "slower" but it is actually doing real work by simulating the folding process itself. (BTW the Linux version is the same version as the Windows client and is as fast as -- if not faster than -- the Windows client.)

In any case, it's good that so many options exist that we can donate our cycles to...

[AnimalMachine]

I also ran the distributedfolding client a litte while back. I stopped using it because of the same thoughts digicosm expressed.

Anyway, I'm on the GLUE bandwagon.

[EatYourGreens]

I have stopped using SETI@home and moved to Folding@home as I feel it is a much more useful use of my CPU cycles.

However, I have a question to ask my fellow folding GLUE users. I have just got a new PC that came with WinXP and now dual boots Gentoo, and this morning I installed the folding@home client under XP. Is it moral/legal/just plain wierd to be a part of the GLUE team while running XP?

[elzbal]

I have just got a new PC that came with WinXP and now dual boots Gentoo, and this morning I installed the folding@home client under XP. Is it moral/legal/just plain wierd to be a part of the GLUE team while running XP?

I'm sure that the GLUE team will appreciate your cycles, regardless of what platform they come from!

[c_kuzmanic]

Cool, just joined the team - I found Seti a little frustrating, maybe E.T. doesn't really use any methods of communications we can detect ?

[Til_Eulenspiegel]

emerge foldingathome fails with the following:

emerge: all ebuilds that could satisfy "foldingathome" have been masked.

Yet the ebuild (foldingathome-3.14.ebuild) shows:

DEPEND=">=sys-apps/baselayout-1.8.0"

and emerge -s baselayout shows version installed 1.8.4.2.

Suggestions on how to emerge foldingathome?

[KiTaSuMbA]

unmasking it? go to your profile.mask and comment out the foldingathome entry.

[henke]

Hmm I'm unable to download the client program. Mozilla and wget stall after downloading 50%-80% of the program Any ideas?

EDIT: Now it works. Weird.

[elzbal]

emerge foldingathome fails with the following:
emerge: all ebuilds that could satisfy "foldingathome" have been masked.
...

If this has not bee figured out by now...
Foldingathome is not masked, but it is currently marked as 'unstable'. Add the following to the /etc/make.conf file:

Code: Select all

ACCEPT_KEYWORDS="~x86"

(Please be aware of the other ramifications of doing this.
I just tried it now, and it emerges fine.

[Mat_le_ouf]

unmasking it? go to your profile.mask and comment out the foldingathome entry.

Hmm, just a stupid question : where is this file?
And I suppose I have to make an emerge sync after the modification...

[Mat_le_ouf]

It was a really stupid question, just had to do an ermerge rsync...
I've installed it and i'm joining the team

[AlterEgo]

I'll keep folding for the Dutch Power Cows (team92)
(I did move my SETI@home box to Gentoo Users )

[klieber]

For anyone that owns or likes TiVo, you may want to temporarily divert your extra CPU cycles to the project described in this thread

--kurt

[EatYourGreens]

After a long and painful process, I have finally got Gentoo up and running on my box. Up to now I have been running the folding@home client under WinXP, but I can now offer much higher quality, open source, unused processor cycles.

And for the record, this is my first forum post from Gentoo.

[arkaine23]

Some things I've noticed in my time folding@home....

The windows client is generally faster than the linux client on the same work units. But if you fold in wine using the win32 client you'll get the same speed as in windows.

The linux client does not idle correctly. It always uses at least 6% of the cpu, and some of the I/O too. Try running openoffice while folding and click on the dropdown menus... you'll notice a lot of sluggishness there... obviously openoffice is too nice in giving up cycles.

In order to correct this so that the rigs at work can all fold, my boss wrote a perl script that STOPs the F@H client when cpu load avg goes up above a set level, and CONTs the client when the load avg drops again. However, since we are now in the process of upgrading the network to gentoo from redhat, we've hit a bug. The FAHcore65.exe shows that it is stopped (T)when we run top, yet it is still grabbing 6% of the cpu and slowing down our engineering software and open office. This is a major pain for us and has slowed down our upgrade (yes we are folding fiends). We think its a problem in the 2.4.19 r9 gentoo-sources kernel, so we're going to try switching back to r5 and hope that the client STOPS when the perl script tells it to STOP. Is there a newr kernel we can try? We absolutely need the gentoo-sources kernel for its free 3.5gb of user addressable ram. Here's a link to the script.

[KiTaSuMbA]

I'm in.... I just emerged FAH. Remember, if you don't generally accept unstable packages you should give the following command:

Code: Select all

# ACCEPT_KEYWORDS="~x86" emerge foldingathome

The go to /opt/foldingathome and edit client.cfg entering your username and the GLUE team id number (11298).
You can then use the init.d/ script to start/stop client or add it to the default runlevel (I didn't make it default yet in order to see if it presents any glitches).
Arkaine says it wont give back all of your cpu but it's related to the kernel you are using. Let's see how this relates to the wolk one I'm using (Currently burning at 100% as nice from what appears on xosview).

[tortoise]

hmm, I'm new to this whole distributed computing thing. Top's telling me F@H's taking up a cool 99.5% of my processor. Is it because I'm not doing anything else right now? Does it take a while to settle down?

It's interesting, at any rate.

[arkaine23]

Folding will use all it can that's available! 99.5% is great. And it backs off really fast when another process needs the cycles. The problems I've hit involved it not backing off all the way. I think I just need to change kernels. 2.4.19 r9 gentoo-sources seems buggy in other things too.

Welcome to the Fold!

www.statsman.org for stats. It may take awhile to finsh a work unit and then a good day to get on the stats.

[KiTaSuMbA]

alright, the -wolk kernel behaves similarly to the gentoo-sources in that respect. F@H won't give it all back. I suspect it has to do with the preemptiveness... I'll give it a spin with a non preemptive kernel and will tell you about it.

[arkaine23]

http://home.swbell.net/trangpho/fold

Here's a perl script my boss wrote. It'll STOP and CONT the folding process based on the cpu load avg. It also switches between the 2 scientific cores if the client sits idle for too long. It worked fine in redhat. But in gentoo, we can't get it to work right. It'll STOP the process, yet the process still uses 6% of the CPU. Very aggravating for us. And we don't know if its a kernel problem or a problem somewhere else. If we don't figure it out, we'll have to rewrite it to kill and restart the folding client instead of STOP and CONT the process.

Seems like 2.4.19 gentoo-sources r5 fixed all our trouble. That and disabling ACPI in the bios. I think my boss has tweaked his script a bit if anyone is interested in the newest version, just ask.

[HunterCo]

Okay, I've got a little server here at home (P2-350, 256 SDRAM) that really doesn't do a whole lot at the moment. So I figured, why not donate it to science, so to speak.
So, F@H is running, and it's all set up to do so on boot as a process, but is there any way to view what it's doing? If I run the console version, I can at least see what's happening. They way it is now, I don't even know if it's working or not...

Update: Well, I know it's running cause it's using roughly 99.6% of the processor... YEEHAH! Now if only I could monitor what it was doing, I'd be in heaven.

[HunterCo]

On my user screen at the F@H website, it lists me as being a member since 2001-09-07, but in reality I only just signed up yesterday. Actually, I never actually signed up for anything, I just threw HunterCo into the USER part of the config file.
Was I suppoed to create an account somewhere first? And if so, am I just giving someone else credit for my WU's?

(addendum added at later time: Now that I think of it, the username HunterCo wasn't taken as I did do a search for it before I downloaded the client... weird)

Is the website just wrong?