This project is "folding protein" in order to find cures, much like Rosetta. Currently they are trying to find cures to Alzheimer's Disease, Cancer, Huntington's Disease, Osteogenesis Imperfecta, Parkinson's Disease, Ribosome & antibiotics. Although it sounds like a good project and works quite well the users in the forums at folding@home don't seem to apricate BOINC as a client due to the connection problems, timelimits, and other issues they believe will not work out.

I on the other hand believe BOINC is a great way to keep computing all the time because I am on dial-up. So if I am off-line for a really long time I would be screwed. Thanks to BOINC I currently have 12 days of work spread across 4 projects plus 60 some days left of CPDN. Willy have you gain any information about folding@home?

The website is: http://folding.stanford.edu/

Currently this program is running almost identlical to seti@home classic. When I started the work unit would have taken me 8 days as a guess. It now dropped down to 4 days. The project was sappose to be BOINC supported mid-January but that data is long gone. If anyone knows more please do say so.

Oddly enough, I read about a third of the BOINC article, and only that article, when I got on their forum earlier today. (Their forum is painfully slow, so I usually only read one or two articles.)

I’ve been running Folding@Home on Windows, Linux, and Mac for several years, and ran it on all my PC’s for a while between SETI classic and BOINC. It is still running on one of my fastest two PC’s at home, and is running on one of the two processors in my Mac. (I have a dual G4, that runs BOINC & F@H. When my new G5 comes in, the two cores will be split this way.) I'm getting very close to 1,000 work units!

If BOINC ever became unstable or had serious issues with all my projects, I’d switch completely to F@H.

There are a few issues. One concern is stability (not just BOINC, but all the shared projects). Another major concern is speed. If you run F@H, that’s usually all you run. For many projects, the work units are released in stages, with successive stages based on the previous results. Since BOINC users tend to run multiple projects, long projects tend to consistently miss the deadlines (think CPN and related projects). I’ve seen “standard” work units take up to eight days on a G4-500MHz processor. The large work units can easily take 4-6 days (sometimes more) on my Intel P4 3.0GHz processor. I’ve never run one of the QMC processes because I’ve only recently had a wintel PC with 1Gb RAM. On my F@H page, it shows number of active processors in last seven and 50 days. The slower processors will sometimes vanish from the seven day count because the work units take 8 or more days (of exclusive 24x7 run time). This is becomming common for my G4.

Anyway, because many of the projects require a steady stream of feedback to progress, if the work units go from two days to a week because they are being shared with an average of three other projects, the F@H projects suddenly can only progress at one fourth the speed.

Another problem is that there are multiple cores. The new Gromac33 is the 8th or 9th core. When you install F@H, you install all the cores.

In a way this last point makes F@H a bit safer, since binary cores, which can be hacked and trojanized, are not transferred. Only data is transferred.

From what I read, it looks like their best option would be to run both clients simultaneously, which would be a pain for tech support. However, this way they can reserve the large and extra large WU’s for their dedicated clients, and have the small and “timeless” WU’s primarily go to BOINC. (The “timeless” WU’s tend to be larger and run longer, but do not have a due date. Users who only fold for a few hours a day are asked to set their system to only get timeless WU’s.)

[quote]
Oddly enough, I read about a third of the BOINC article, and only that article, when I got on their forum earlier today. (Their forum is painfully slow, so I usually only read one or two articles.)

I’ve been running Folding@Home on Windows, Linux, and Mac for several years, and ran it on all my PC’s for a while between SETI classic and BOINC. It is still running on one of my fastest two PC’s at home, and is running on one of the two processors in my Mac. (I have a dual G4, that runs BOINC & F@H. When my new G5 comes in, the two cores will be split this way.) I'm getting very close to 1,000 work units! And that's a tough task!

I did some test run some time ago with the boinc version of folding, but now the site is completly gone - strange...

- Peter

Not that strange, just a small peek at their communication skills

I didn't realize F@H downloaded binaries. I thought all binaries were self-contained in the install. My mistake.

If F@H goes to BOINC only, there will have to be a way to allow for their four options of work units, [1] timeless only, [2] standard, [3] large, and [4] very large (QMC units, possibly unstable). I run standard at work and large at home on my Intel. I've read that the very large units are resource-demanding, and also are the most likely to crash. The only way I can imagine this working is something like what Climate Prediction does. You can register for the base project, which has large work units, or you can specifically register for the much larger sub-projects, and they ask you to please please please do not run them concurrently.

There are seveal articles on F@H about points. They count both work units and points. They say that when they give points a positive adjustments, they have old timers threatening to leave because their years of credit is cheapened. If the points favor large WU's and lower standard ones, the new users get discouraged and threaten to leave.

Of course, I'd love to ask one of these point junkies, "And just what the hell do you expect to do with these points?!? Do you get free long distance minutes through your cell phone or something?"

<Off Tangent Alert>
That's why I'd like to see actual work units in BOINC results. For F@H, I'd be quite happy if my score was total work units, then four scores for the four different kinds of WU's. (A secondary breakdown by core could also be cool.) Same way in BOINC. If I see someone has 100 SIMAP WU's complete, that's nice, but not special (5-6 days on a P3-1GHz). If I see 100 FightAIDS@Home units, that's several months work. If I see 100 CPN units, that's really impressive.
</Off Tangent Alert>

They stated quite often, and not only the Forum-mods but Vijay himself, that they will never switch to BOINC-only, and that's part of their difficulties. But I think, if funding is OK (unlike Seti), it's the better way for not upsetting the veterans.


If BOINC will manage to account for the true crunching amount one far day, it'll be the stop the bickering IMHO. WUs are far to different to be counted as just that, if you're in the competition (and that's a good incentive for many to join).
FAH points are fine as well, although I personally don't like their bonus system (additional points for big WUs, or big DL/UL, or whatever else). They are a different amount, but work similar to CPDN trickles (predetermined points per WU).

Points are a good incentive, but the competition gets annoying to some. Personally, it sets my butt on fire when someone hacks the BOINC manager to get lots of points (and usually ends up returning bad results), or hijacks another's account to steal the points. We're contributing to science and engineering (and a few projects for art and liesure), but people will ruin the incentive.

Really, like I said earlier, does your long distance phone service give you free minutes or something? (I'm speaking to no one here. It's a general complaint, not to anyone on this thread.)

I know that different project work units are not the same. I run SIMAP (~32 minutes per WU on an Athlon XP 3200+) and FightAIDS@Home (~12 hours per WU), and not everyone would know the difference in value.

Perhaps something like you said, Sänger?

Instead of points, the BOINC manager could give (1) WU's crunched and (2) total crunching time? This is how SETI classic and the old Entropia FightAIDS@Home clients worked. For both, you would see total work units and total computational time for all clients. (The comp time feature was broken for the last two years on the Entropia client.) The only problem is that doesn't give "total computational power". Still, I could say with pride that I donated 12 years time to SETI@Home. On a human level, that "feels" more "real" than saying, "I have 4,500 points for SIMAP." A person could also argue that 12 years comp time on a Pentium 2 and Pentium 3 is not nearly the same as 12 years on a P4 or Apple G5. Then again, you would have a very different total number of work units and average timer per WU....

It's too late to give wu count and actual crunching time. Oh, and btw the boinc credit system is already widely used to calculate computational speed and in some cases to calculate the time a 1GFlop computer would have used to do the same amount of number crunching. So, dropping the boinc credit system (aka the copplestones) would bring the whole boinc community down to it's knees

Not to mention I would have to find another hobby.

Basically what I also meant with 'bring the whole boinc community down to it's knees' All the time and efford by you and everyone else who use the credit system in various ways would be rendered null all of a sudden - I don't think many would like to start all over from scratch not even being able to produce the same or equal result (and I personally would have saved more than 6 months worth of XMLs for nothing).

[quote]
Basically what I also meant with 'bring the whole boinc community down to it's knees'

Anyway... And now for something back on topic.

I was reading a BOINC thread on the F@H forum last night. It looks like they are talking about a fork in the clients. The article said the BOINC client would be good for small work units, timeless WU's, and WU's that otherwise have long deadlines. As for the large and very large WU's, they would stay with the classic client, which usually runs independantly of other projects. It sounds almost like new studies would be screend to "BOINC-friendly" and "Classic-necessity", and their work be written for one platform or the other.

Ok, sorry if I miss-interpreted your motives for your post. Anyway, this all goes back to a point about BOINC that, as I see it, has flown right over the heads of people, pardon my expression

Which is also why I hate cheaters. Personally I couldn't care less about people wanting to boost their rankings by cheating (if it wasn't amoral), but it affects both project wide and BOINC wide calculations, like speed, years of computation etc.

Anyway, I was worried that they would scrap the idea of supporting the BOINC platform when the beta site disappeared - so what you're saying here is good news indeed. Although I wonder about the wu's they think don't work with BOINC??? I mean, if they don't work with BOINC they propably don't work with any client running on average users computers, or what is their explanation?

That's OK. I wasn't trying to derail the conversation. Actually, I was reading about scores in the F@H forum today.

Some F@H studies are time sensitive. The work units are released with the initial amino acid chain, and find the first few folds that are lowest energy states. They are sent back, and another batch are set out with the most promising results. These fold the next few steps, get sent back, and analyzed. The third cycle come out with the first few steps of folding at a lowest energy state. This continues a few dozen times until something is folded, hopefully the correct protein shape. So, because of this, some studies are time sensitive. A quorum has to be reached for each step. If each step is delayed for a few days, the project can be held up by months. The faster the results come back, the faster the research progresses. That's the time problem.

The size/stability problem comes with some of their large and very large work units. The large units can take a week or more on a P4-3GHz with 24x7 runtime (5-8 days seems the normal range on my P4-3.0GHz). Again, some of these are time sensitive, as above. If the units already take a week on a high end PC, and there are incremental steps in the research, having the time for each step doubled can slow the research down dramatically.

The very large units (I think "QMD" units) require at minimum 1Gb of RAM, preferably two. Also, the recomended PC is a high end P4 or equivalent. The run time is several weeks. Further, since these are experimental, they are not as stable as normal work units. The FAQ says that they should only be run on high-end PC's that are checked regularly and can be easily administered. Just like the various CPN spinn-offs, it is "recomended" that they be run without other projects, and will take forever.

I've been running F@H for a few years, and I've run both regular and large WU's. Ironically, the three times I've had a problem have all bee on my Mac. (I could restart F@H, and it would return to the last checkpoint, only to crash at the same step. In all three cases, deleting the work unit fixed the problem.) So far, I have never had a large WU crash on my Intel PC.

the problem with WU count and CPU time is that it's much harder to compare users/teams/projects etc. to one another, yes the boinc system isn't perfect, but what system is?
it's the ideals/principals that matter along with progress toward achieving them, which is exactly what's happening, i personally perfer the boinc credit system as the "master" value

however, on the other hand it would be nice to know what my 100,000 or so credit means in "old money", that is CPU time and WU count, but not use it as the "main" credit value

I am still of the opinion that by fully utilizing the BOINC infrastructure F@H could send any type of task to a BOINC participant. The powers that be on the project do not believe that and the way they are currently implementing the BOINC version their expectations will be fulfilled.

My real hope is that during this extended outage of the F@H BOINC beta they are rewritting the code to take better advantage of BOINC. We have been telling them that they need to almost from the beginning, hopefully they have been listening.