Packaging F@H in an MSI - Allowed?

[Shuriken1]

Hi Everyone,

I'm new here although not so new to the F@H project as a whole. I'm a network manager at a school in the UK and it struck me recently that schools have lots of idle computers. I've managed to compile the console version of the 'classic' client (v6.23) into an msi that installs it as a service, etc etc. It's all working nicely and will be nice and easy for school IT staff to install - I'm hoping to promote it and get a bit of a school Folding@Home community going on.

It's struck me though, I might be in violation of the licence agreement. I read through it and it does clearly say that it must only be obtained from the Stanford site, but I was wondering what their stance might be given the educational link. The msi files would be tailored to each school and not publicly available. I was looking to make them only install onto PCs in a certain domain so that the msi's could not be misused, to better meet the licence agreement, although didn't get very far.

I was wondering whether this has come up before and whether any regulars (or even admins) might be able to give me a more enlightened (or official) response.

I'd really love to get Folding@Home prevelant in schools in the UK, but I realise I would technically be breaking the licence agreement. I really hope someone from the F@H team might be able to consider this case and grant me some exception - althought I realise that's probably unlikely!

Anyway, any thoughts would be very welcome.

Shuriken1.

[bruce]

Welcome to foldingforum.org, Shuriken1.

An "official" answer can only come from the Pande Group, but the question about redistributing FAH has been asked a number of times and the answer has always been NO except for actual development partners like ATI, Nvidia, Sony, etc. (In fact, the PS3 client is distributed by Sony, not by Stanford.)

Some 3rd party developers have (briefly) distributed code that included the Client and Stanford has aggressively insisted that they revise their code so that it downloads the client directly from Stanford.

I agree with your assessment: It's unlikely that Stanford will grant you an exception.

[Shuriken1]

That's a shame.

I'll hold out for an official answer (hoping they have time) but I think I know what the answer will be. Although, if I can configure the msi to download the fah .exe as part of the installation process rather than including it in the msi file, would that make it allowable? I don't know that I can do that, just thinking out loud really. I really wanna make something of this project!

(Obviously this is more a question for Pande representatives now although I'd still welcome other input)

Thanks for your reply Bruce.

Shuriken1.

[MtM]

Welcome to foldingforum.org, Shuriken1.

An "official" answer can only come from the Pande Group, but the question about redistributing FAH has been asked a number of times and the answer has always been NO except for actual development partners like ATI, Nvidia, Sony, etc. (In fact, the PS3 client is distributed by Sony, not by Stanford.)

Some 3rd party developers have (briefly) distributed code that included the Client and Stanford has aggressively insisted that they revise their code so that it downloads the client directly from Stanford.

I agree with your assessment: It's unlikely that Stanford will grant you an exception.

I would like to add that even downloading directly from Stanford is prohibited unless written permission is obtained beforehand.

At this time only fInstall and inCrease have this permission afaik, when I lobbied for written permission for FahTracker it got denied probably because of the closed source nature of that project.

If your project is open source, one can always try. But it will not be an easy task, you will have to convince them of the good nature of your application and that it serves a purpose ( eg offer something the official client distributions can't do ).

[Shuriken1]

Ok, thanks for your reply MtM, I'll keep it in mind.

Shuriken1.

[Brazos]

Little off topic but bear in mind that the use of folding@home on pc's will increase the electricity use of the school. There was case here in the US where a IT person put multiple instances on his schools servers without the knowledge of his superiors and he was fired. Using public funds to do folding without permission may not be a wise thing to do. My 2 cents.

[Shuriken1]

Hi Brazos,

Yep, fair point. I was also concerned about the state of the machines, so I decided that I'm only going to run it at about 30-50%. I think that should keep heat, damage, electricity costs etc down. 30% is pretty low but the way I see it, 30% of 300 computers is better than nothing! Should be alright.

Shuriken1.

[Shuriken1]

Oh, and I'm not gonna put it on the servers, that properly would hike our bill up a bit! lol. Just the pc's.

Shuriken1.

[MtM]

Hi Brazos,

Yep, fair point. I was also concerned about the state of the machines, so I decided that I'm only going to run it at about 30-50%. I think that should keep heat, damage, electricity costs etc down. 30% is pretty low but the way I see it, 30% of 300 computers is better than nothing! Should be alright.

Shuriken1.

You should note that the deadlines for even classic client work units are pretty short right now and not knowing the hardware involved running 8 hours a day for 50% might not be enough.

If you post their specs and their normal usage pattern it would help narrow that down allot!

[Shuriken1]

You should note that the deadlines for even classic client work units are pretty short right now and not knowing the hardware involved running 8 hours a day for 50% might not be enough.

Oh really?!

Most of our computers are dual-core ~2.8Ghz, 2 or 3Gb of RAM, HPs of various description mostly. They're pretty decent as classroom computers go, but then if this were to go into other schools, who knows. Part of the reason of not running it at 100% is so it doesn't burn out some old PCs schools might be running that I don't know about. It's a bit difficult to judge really 'cos the SMP client sounds like the WUs are too short but the classic client only works on one core and I don't know how many cores this msi might be installed onto. Hence sticking with the classic client at 50 ish percent - play it safe!

I started the client on my PC in the office and granted that runs 24hr but the due date on that was the 5th of July or something and it had the best part of it done overnight. That time spread out over working hours would still be done by the 5th of July I reckon. Do the WUs have different time deadlines though? I'm presuming they're all the same..

Appreciating all the input from people, by the way, thank you. I'm used to posts just being ignored on forums, makes a nice change.

Cheers,

Shuriken1.

[MtM]

http://fah-web.stanford.edu/psummary.html gromacs and amber are classic core's.

You can install the classic client as many times as you have cores, which will use 100% cpu time. If you want to cut that down, use one client per two cores and let it run 100%, the same result but the work unit is returned much quicker and that's preferred. The science benefits more from faster returns then from more returns, go read around the forum ( search for Quick Return Bonus or QRB or dual smp if you're interested ).

My E2160 does wu's with plenty time left ( core duo 1.8g ), but run's 24/7. 2.8G doesn't say anything without the architecture. If you're using c2d's or higher I wouldn't worry, but still test it on one machine first before rolling out over the entire school.

[Shuriken1]

You can install the classic client as many times as you have cores, which will use 100% cpu time. If you want to cut that down, use one client per two cores and let it run 100%, the same result but the work unit is returned much quicker and that's preferred.

Yeah, the problem is that if this msi were rolled out across a school (or any organisation), they could be some dual cores, some single cores, maybe some quad cores. I've heard that about WUs being done quicker is better than lots being done slower, that's cool. Just difficult to work out a good balance with an unknown number of cores. I could always see if the msi can read the number of cores and install that many copies, but that's starting to sound complicated!

I'm not sure what architectures of the top of my head, I'm not sure we've got many C2D's, most Pentiums of the EXXXX variety I think. Again, more an AMD man so not up to date on Intel chips, just go by the Ghz.

I can't believe that a dual-core 2.8Ghz running 8 hrs a day couldn't meet the deadlines but, I'm only a newbie so what do I know! And yes, I wasn't just gonna dump it out across the school and pray, lol.

Thanks for your help MtM.

Shuriken1.

[MtM]

Sorry I had a reply ready but got distracted and closed the tab unknowingly.

In you case I would still install a single classic client on a machine representative of most others, either with v6 or v7. If you choose the default setup for uniprocessor using V7 that would be enough. Monitor it, with v6 use an external monitor like fahmon or hfm.net ( look in the 3rd party forum section ) and check if you can make the deadlines! You could try that with an smp client / slot as well. If you're making deadlines on that machine, it's safe to assume you will make them on the other equal machines considering their equal time switched on.

It would be a shame if you installed clients on lots of machines without knowing for sure they will make the deadlines, as not only your contributions wouldn't count as they are given out to other donors after the deadline expires, but you would also waste your higher electricity bill for nothing.

[codysluder]

I recommend you follow the KISS principle and it simple. Why not just install one uniprocessor client on every machine. Leaving it at the default 100% CPU setting, it will use 50% of any Dual-core hardware and less if there are more cores. That's not maximizing performance, but it is easy and it's still quite a bit of folding. The client will have zero impact on normal classroom use.

Maybe next year, consider refining it by counting cores and customizing the installs.

Since you're an AMD man, be aware that many Intel chips use Hyperthreading. The hardware creates virtual CPUs which can interleave processing from two tasks, allowing for better instruction reordering. There are some extra integer registers but they share one FPU so they don't double the performance of scientific calculations. By running one client, you don't have to decide whether the hardware has Hyper-threaded cores or not.

[Shuriken1]

Hi codysluder,

Yeah, that's pretty much my theory at the moment, other the fact that I was planning on playing it even safer with one uniprocessor client at 50%. What makes sense though is to just tell schools not to install it on single-core PCs! Not sure why I didn't think of that before, but that's definitely abiding by KISS principles. I think I'll do that, 1 client at 100% and tell them to only install on dual (or more) core machines.

Thanks for the Hyperthreading info too, I've always known that HT is kind of 'fake' but never known much more than that. The fact that they share one FPU I didn't know so that's cool, thanks.

Just hope someone from the Pande group looks down favourably on my little endeavour now and doesn't stop this project before it starts! On that note, am I likely to get a response from any official person, or do they not frequent these forums often? Not trying to suggest they don't, just don't know how things are around here.

Thanks again all,

Shuriken1.