TR Forums

cass · Sun Oct 10, 2004 11:24 am

Very good post Emkubed, and it comes back to something that I have asked already.... Who is going to be considered the owner of the stuff?

I personally think we are going to need the cooperation and input from the TRsite honcho's. This is their Brand, reputation, and Business we are putting out there.

I think the stuff we donate should be considered TR property loaned under contract. Once and when this is over I kind of think It should be liquidated and proceeds sent to TR.

We do need a Treasurer and If Emkubed is game I am fine with it. However, I think the Paypal account needs to be Owned by TR and set up by TR. Therefore there will never be a dispute over money if any accumulates there.

As far as making purchases go, I think the treasurer will need to power to buy at will what is needed, but think there needs to be some loose general guidelines. For example just because FX53's are on clearance don't necessarily mean we need a $500 proc. I think $100/proc is enough, and $50 for a MB is reasonable.

Just some Ideas.

*emkubed PM me with 25% amount.

idchafee · Sun Oct 10, 2004 11:28 am

cass wrote:
I personally think we are going to need the cooperation and input from the TRsite honcho's. This is their Brand, reputation, and Business we are putting out there.

We do need a Treasurer and If Emkubed is game I am fine with it. However, I think the Paypal account needs to be Owned by TR and set up by TR. Therefore there will never be a dispute over money if any accumulates there.

I'm good with emkubed being the treasurer or whatever. TR has a Paypal account already. Now, whether Damage wants to tie that acct in with FBot is another matter altogether.

cass wrote:
*emkubed PM me with 25% amount.

I'll take 25% too. I know the paypal addy, just need the amount.

cass · Sun Oct 10, 2004 12:18 pm

I'm thinking a Separate Paypal acct for Fbot would be better.

Sun Oct 10, 2004 6:39 pm

I had a VERY long reply that my browser ate right before it died, so I'll try to recreate it succinctly.

I don't know that asking Scott and Geoff to take more of their time up by monitoring an account that shouldn't get a lot of activity, and then having to act as middleman to dole out those funds for orders, would be prudent.

On the issue of ownership, though the donated FB parts belong to the team, and the team itself is free (as in voluntary, requiring no registration, obligations or responsibilities), the hosts are using their electricity, space and time to shelter these machines. Now, the project is going to continue as long as there is a TR, and as long as Stanford continues the project, but either way, by the time it ends, these machines (looking at the hardware today) will have quite a low actual value. I think we all agree that machines made solely for the advancement of the team, built by donated time and parts, were built for TR, and if it all comes to an end, the owners of TR get first say. But if the donated parts are by then throw away spares, I'd say the hosts may well have paid for them in electricity.

Granted, the project's donated parts are going to change with the times, but I think if we keep public account of who hosts what, and acknowledge that the foremost owners are Scott and Geoff (if they want 'em), everything else will take care of itself.

That was a little less well-put than my first try, but that's the jist of it.

I'll email the guys to get their opinion on ownership and accountability.

cass · Sun Oct 10, 2004 7:04 pm

I don't think they (TR powers) need to have to worry about running the TRbot account other than being the ultimate owner and controller. I mean the Treasurer/secretary will be the one doing that, but TR powers should have the power to come in and lock down the account over that. Hopefully TRpowers would never have to do anything but watch the sideshow.

Good point on the equipment, I surely don't plan on seeing anything I put in again.

All this is merely meant for documentation so is someone gets peeved/banned they have been put on notice the equipment is not theirs. Although it is unlikely anybody would want it other than for some DC proj.

Usacomp2k3 · Sun Oct 10, 2004 7:21 pm

sounds good to me emkubed, although I'm not a host or anything...some of the first parts ever came from me

leor · Sun Oct 10, 2004 10:03 pm

at this point, people like us that have been around foever i think can be responsible to traffic ourselves.

none of us are going anywhere, and we put a lot of our pare time into this site, whether it be from writing, moderating or pixel pushing. We may not actually be staff but we're certainly close enough to the point where we can handle the organization of this project on our own.

I also agree that in the event that either TR or stanford's project ends, the hosts will by then have paid enough in electricity to cover the cost of whatever boxes they have been hosting(since i imagine TR isn't going anywhere and we are unlikely going to unlock all the secrets of the protein in the next year or 3).

go yankees :evil:

Sun Oct 10, 2004 10:42 pm

I agree with emkubed, we do no want to request that the TR staff assume any additional administrative responsibility. While the efforts of the TR F@h team do add some prestige to the TR site, they are an independent effort, and should remain independently managed unless TR staff wants to get involved.

I believe that the FrankenBot hardware is collectively owned by all people who have contributed hardware, money, electricity, or other resources to the FrankenBot effort. However, it is impractical to accurately compute how much any given contributor owns. Therefore, anyone contributing hardware or other resources should consider it to be a donation.

If a host is no longer in a position to host the system(s), see if someone else on the team is willing to take over the hosting. If a FrankenBot system is being retired (instead of being rehosted with a different team member), pump the proceeds (if any) from liquidating the hardware back into the FrankenBot project.

I would also like to take this opportunity to offer my services if anyone has a motherboard with bad caps, which could be a worthwhile addition to the FrankenBot project if repaired. I've "recapped" quite a few mobos, with a pretty decent success rate. If you have a failed mobo with obvious signs of bad (bulging/leaking/exploded) caps, I'd be willing to attempt to repair it free of charge, with the stipulation that it will be used for folding.

Tue Oct 12, 2004 12:46 pm

Damn, I forgout about your re-capping experience. We'll definitely have to use you on those old socket A boards!

cass · Tue Oct 12, 2004 11:09 pm

TRFrankenBot Diskless #1 is now running.

256M 2100 2.5 ram
AXP1600 tbred proc. stock speed

P1308 one percent every 32min. 144 pts/day on large WU

137M memory active used, 53M memory free
rest is cache, inactive.

Wed Oct 13, 2004 4:57 am

cass wrote:
TRFrankenBot Diskless #1 is now running.

256M 2100 2.5 ram
AXP1600 tbred proc. stock speed

P1308 one percent every 32min. 144 pts/day on large WU

137M memory active used, 53M memory free
rest is cache, inactive.

Not sure if the caps matter, but it's TRFrankenbot, without the capital B.

w00t! If Gandolf will up his bot, we'll go up to 5 CPUs this week!

cass · Wed Oct 13, 2004 8:43 am

Thanks, just double checked, I have it right in the client.cfg.

I was just tired when I typed it here.

idchafee · Wed Oct 13, 2004 8:45 am

emkubed wrote:
w00t! If Gandolf will up his bot, we'll go up to 5 CPUs this week!

Uh, Gandolf? Haven't you had that bad boy for a week now? Lets get it turned on!

cass · Sun Oct 17, 2004 9:55 pm

scrounged a couple old hdd's out of the pile. wired me a board up and started configuring FreeBSD 5.2.1. This will not be a board that I donate, but I can go ahead and get the system and disk images built. Got updated sources and fetching ports now, will try and build kernels tomorrow.

1st diskless node is permorming nicely. No lockups or dumps.

cass · Thu Oct 21, 2004 10:46 am

Got a 6.4G hdd and finally after several attempts got freeBSD 5.2.1 loaded and got a custom kernel with sse compiled, but there is some type problem. The system is running at 1/4 speed or worse. Everything looks OK from bios level, but system is just way slow. I don't know but will keep plugging away.

I can start the diskless server setup anyway, I think the slow part is either a driver or device that I don't have in the kernel. The default kernel runs the same way.... time to check for nforce 2 support @ freebsd.org.

cass · Thu Oct 21, 2004 5:29 pm

Just received 1 Motherboard... some type with 512M pc133 and tbird? proc. with nice aluminum eargersplittenloudenbacher turbo tower fan. It appears all this baby needs is a junk vid card... I happen to have one... well the cooler bottom was machined with a roughing tool and gouged some, but I will remedy that :wink:

.

This will be running tonight late. Frankenbot .... mush

There is also an nforce2 MB w/two video outs and lan in the box.... I have a proc in mind for that one....

I don't know where this came from Idchafee or emkubed, but Its here.. It will be running before long.

Thu Oct 21, 2004 5:31 pm

I sent you 2 mobos and that rockin IBM managed switch. That might be package 1 of 3.

I also have Converts mobo's and RAM I'm going to send your way for mounting in the new diskless farm.

idchafee · Thu Oct 21, 2004 7:46 pm

emkubed wrote:
I sent you 2 mobos and that rockin IBM managed switch. That might be package 1 of 3.

I also have Converts mobo's and RAM I'm going to send your way for mounting in the new diskless farm.

Will any of those mobo's hold a celeron? I have a celeron 2.0 ghz. and a buttload of HDD's that I imagine cass can't use.

cass · Thu Oct 21, 2004 7:59 pm

find out what type MB that celeron needs.... socket, chipset etc.

I'm sure I can find something on ebay.

well, what type/size hdd's are they?

We can use up to 4 drives if need be, but I think 2 or three would suffice.

currently have 1. I am using couple of mine to get going. the 6.4G drive is staying with the bot. I am afraid of the 2.1G and 406M I have because they have already proven they will not run long.

cass · Fri Oct 22, 2004 7:14 am

Minor setback here. FreeBSD 5.2.1 is an absolute piece of crap for doing anything but trouble shooting someone elses **** up's. I have never got it to use over 40% of cpu, and of that small slice, 60% of that is running interrupts. I have tried several kernels, which was a grand waste of time because it took 8hours to compile one. That is a bit excessive being it takes like 10 minutes in FreeBSD 4.8. Both machines same processor.

I got a sse kernel going, and for kicks started folding... I ended up getting a 400 frame p1130 which is tinker. It ran for 8 hours and still did not complete one FRAME.

The motherboard is Nforce2 and cruising BSD mailing lists I can see some people think they are getting good performance. Of course I also remember Nforce1 and being called an idiot for over a month by others convinced there was nothing wrong with the pci bus.

The "developers" continue to deny there is that big a problem and point to 5.3 so, tonight I will try 5.3. I don't think it will be any better, but I will try it before reverting to 4.10.

I put a boot nic in the nforce2 MB, added it as a node to my diskless farm, and its doing great. one frame every 6min15s for the first two.

I did read some noise about dragonflyBSD developers leaving freeBSD over the future path. I guess some laid off MS developers moved in to help the FreeBSD boys.

Fri Oct 22, 2004 8:03 am

Yechh...

Time to give Linux a try, perhaps? :wink:

cass · Fri Oct 22, 2004 8:28 am

It will be, if 5.3 is no better, and 4.10 does not support amd64 properly. I don't like the idea of linux's canned kernel, but I could probably make it work.

Fri Oct 22, 2004 8:34 am

cass wrote:
It will be, if 5.3 is no better, and 4.10 does not support amd64 properly. I don't like the idea of linux's canned kernel, but I could probably make it work.

Well you're certainly still welcome to tinker with the kernel if you wish; the point is that you generally don't need to, for something as simple as folding. I run stock Redhat/Fedora on all of my disk-based Linux systems, and stock LTSP on all of the diskless ones. :wink:

And while Fedora is available in an AMD64 version, I don't think it really matters, as the folding client is 32-bit only.

cass · Fri Oct 22, 2004 3:25 pm

emkubed wrote:
I sent you 2 mobos and that rockin IBM managed switch. That might be package 1 of 3.

Got the switch.... NICE. How in the world does something that nice get retired.? Gigabit?

Yall are gonna make me have to buy lumber now .

Fri Oct 22, 2004 3:43 pm

cass wrote:
emkubed wrote:
I sent you 2 mobos and that rockin IBM managed switch. That might be package 1 of 3.

Got the switch.... NICE. How in the world does something that nice get retired.? Gigabit?

Yall are gonna make me have to buy lumber now .

That's a hell of a switch, isn't it? Well, at least when troubleshooting network issues you won't have to worry about the switch being a problem.

How many nodes does it take for you to fill up a new cube full o' Folding goodness? We could supply a 12 port switch per cube.

cass · Fri Oct 22, 2004 4:24 pm

I am kicking around 2 sizes right now.... 1 server 5 diskless for 6 total or 1 server 11 diskless which would fit 12 port switch fine. The 6 node would be rounghly 15x14x24 and the 12 node 15x14x48. I am probably going to use cheap 5/8" underlayment chipboard with 1x2 or 1x3 around the outside corners for strength. Of course a 6 noder with an intergrated 8 port switch would give room to daisy chain boxes togither with an external port left for someone to plug in another system. 12 noder would have no outside connection available except to plug up Fbot. Uness the 12 ports switch was a 12+1 deal.

I currently have the server setup with two interfaces. The interface that goes out to the wan responds to nothing, no rlogin, no ssh nothing. The one on the internal lan is natted to the outside and then the outside is firewalled. I don't really know what I am doing, but have some examples and tutorials to help with Fbot. The Lan card plugs to the switch. so anything that plugged into the Fbot switch will be firewalled and reasonably safe. The Wan card is the only one that will appear to the local network where Fbot is installed. The Fbot internals will be invisible.

I found out what is wrong with my BSD install after much reading. there are three separate problems, scheduler conflicts, gcc 3.3.3 screwing up floating point, and some driver stuff. Maybe 5.3 will get around some of it, but at this point 4.10 would work. 5.3 would be more better from a point that in a year it will still be supported. 4.10 will probably be abandoned around a year from now.

I will go on over to networking now for my next question.

cass · Fri Oct 22, 2004 9:21 pm

Success with freeBSD 5.3.... well OK it runs and folds faster than any windows box I have. Make that faster than anything I have period.

In a project 1130 it is doing a frame every 6min 0s the next best thing I have is 6min15s with the same processor.. axp2400.

This is with the generic kernel. I am able to compile a custom kernel in under 5 mins now.

cass · Sat Oct 23, 2004 9:42 am

Well now, I have FreeBSD 5.3 running, with all new sources, custom kernel compiled and running... It actually takes 10 minutes to do a full compile of sources and all. Have Xorg installed, and Gnome. Usermin is installed too. Have ran CVSUP and did a buildworld too..... first time I have got everything to work.....

I did have some type problem with a bad cd I think... I could not get the GENERIC kernel to compile, so I CVSUP'd in new sources and everthing is groovy.

cass · Sat Oct 23, 2004 12:26 pm

One thing that is bugging me that I have no solution to yet is how I am going to delay the start of the diskless nodes untill the server is running.

If I set everything to boot on power up, the pxe of the dislkless nodes will timeout before the server boots and gets dhcp started.

I don't know how much trouble wake on lan would be or if it would even work.

I really need a one minute delay between start of server and start of nodes then it would jsut be simple with one button.

Sat Oct 23, 2004 12:30 pm

Can you have a script that runs on the server at the last runlevel that sends a WOL to each node?

TR Forums

Frankenbot rebirth. The cass farm of DOOM.

Who is online