TR Forums

Just want to know what is the SMALLEST size USB stick I can use for folding (size in MB)? I have a 128 MB USB stick, and a 512 MB USB stick; would either work for diskless folding? If not, what is the smallest size USB stick I can use?

Also, what is the minimum amount of RAM I need for diskless folding? I'm trying to put an old 2.0 GHz P4 to good use by using it as a diskless folder, but as a senior citizen on a VERY tight income, I need to keep costs to the absolute minimum, within reason... I have a few other slower boxes I can deploy (five more P4s, ranging from 1.6 GHz to 1.8 GHz), but they all need to be assembled, and have various other hardware issues yet to be resolved (need suitable RAM and PSUs, along with two more HSFs), so they won't be ready to go into service for a while yet...

Those systems were all donated to me a few weeks ago by the owner of a small business (he worked alone), who recently retired due to health problems. I promised him that I would not simply allow the hardware to sit and rot, so I want to get them working as diskless folders, as time and (very little) budget allow...

Thanks for your input; I don't know ANYTHING about Linux yet, but that will change VERY soon. I'm typing this on a friend's Win2k box in mild weather, while I avoid the effects of Old Man Winter back in New England... I want to be able to hit the ground running when I go home, and get those systems folding with USB sticks, so any and all advice is appreciated.

My rig at home runs XP Pro, and is folding (for a different team) in memory of a friend who was a dedicated folder. He lost his battle with cancer, but his widow told me about the Stanford folding project, so I joined his former team (so I could keep his account active), and have been folding ever since...

Lastly, my thanks to notfred for his incredible work on this project, and to all the other people willing to teach retired coffee addicts how to help in the ongoing fight against disease... Once I know what size USB sticks will work as diskless folders, I'll be able to put at least one system online fairly soon, with the others to follow...

Fold for the cure...

With the USB stick it is used for booting and for backup. The boot stuff is only 16.8MB so the big question is how big is the backup and that will depend on what WUs you accept. I think you will probably be OK with 128MB, but if you want to be safe the 512MB would easily do it. I don't keep too much of an eye on what the backup size is, some others here may have more measurements.

Welcome to the team. I love your passion and your ambition. Hopefully we can do everything we can to take advantage of your needs

@notfred:

Thanks for the information about suitable USB stick sizes; I'll experiment with the 128MB stick, and invest in a few 1GB USB drives, just to cover the bases... BTW, notfred, do you happen to know the minimum amount of RAM required to make this work? I get the feeling that just about ANY size DIMM will do the job, but if you happen to know that a specific amount of RAM is required, I'd like to know what to budget for. I get the feeling I'm going to be doing some RAM experiments to see what effect using various sized DIMMs will have, and whether there is an improvement in work unit production by increasing the amount of RAM to a certain minimum threshold... What the hell, I should have some free time to research this, if no one else has had the opportunity yet...

Now, a question about some instructions from this webpage:

http://reilly.homeip.net/folding/usb.html

The instructions in question are:

"Edit the file isolinux.sys and change the USER and TEAM options to get it to fold for you rather than me. You may also want to change BIG=yes..."

I've downloaded the USB folding files (usb.zip and syslinux.exe) and examined them with 7-Zip; I didn't find any file named "isolinux.sys"... Is the actual file to be edited named syslinux.cfg? I opened it with a text editor, and (except for the different name) it seems to match up with the instructions; if those instructions do NOT apply to syslinux.cfg, can you point me in the right direction on locating a copy of isolinux.sys? If it is a text file, you can paste it here, and I can copy it into Notepad, and create a new version without having to download anything...

Other than that, it looks like I'm going to have to learn something about networking while I'm learning Linux... I've already downloaded and burned a few .ISOs (Puppy Linux and PCLinuxOS, just to name two) so I can start to get familiar with using Linux before I go back to New England...

Thanks again for all you've done to bring diskless folding to the clueless masses (well, ME, anyway), notfred; I'm looking forward to the day when I can FINALLY get my first diskless folder online... to be followed by others...

@Usacomp2k3:

Thanks for the welcome; I won't actually have a diskless system online until late spring at the earliest, but by learning everything I can now, I hope to be ready to start a diskless folding farm the day I get home. In the meantime, I've got a pretty steep learning curve ahead of me...

You're right, it should be syslinux.cfg and not isolinux.sys - I've fixed it now, thanks for spotting that.

I think any DIMM that you have for a P4 will probably work - you'll start getting nasty messages on the console if it runs out of memory.

coffeecrazy wrote:

@Usacomp2k3:

Thanks for the welcome; I won't actually have a diskless system online until late spring at the earliest, but by learning everything I can now, I hope to be ready to start a diskless folding farm the day I get home. In the meantime, I've got a pretty steep learning curve ahead of me...

You can probably start by playing with virtual machines. I believe those can do PXE booting.

I've found a 128 MB stick is not large enough to consistently succeed with the SMP client, but it will probably work with the others, though I've not verified that. I simply know that when running notfred's SMP client, I often get write errors because the 128 MB stick is full. Sometimes I can move the USB stick to another machine, delete everything on it, and put it back and it will start working again if I notice the error (which appears to only show up on stderr so you can't tell it's happening unless you've got a monitor connected to the machine) before it's completed its final frame. Otherwise I have to hope the thing will finish successfully and send the WU, though I think one could copy the WU backup from the web page if the thing failed. I often forget to do that when I discover a problem, though. I lost 2 WU's that way in the last week on one machine using the February 2 diskless client, but that's a discussion for another thread.

coffeecrazy wrote:

Also, what is the minimum amount of RAM I need for diskless folding? I'm trying to put an old 2.0 GHz P4 to good use by using it as a diskless folder, but as a senior citizen on a VERY tight income, I need to keep costs to the absolute minimum, within reason... I have a few other slower boxes I can deploy (five more P4s, ranging from 1.6 GHz to 1.8 GHz), but they all need to be assembled, and have various other hardware issues yet to be resolved (need suitable RAM and PSUs, along with two more HSFs), so they won't be ready to go into service for a while yet...

Your P4s will be limited to single core work units. You should be fine with 128MBs of RAM and might be OK with 64MBs. For maximum points/science production you need at least 256MBs of RAM to qualify for the "big packets" (double points bonus) work units. Although lately the assignment server has been ignoring any machine that doesn't have more than 256MBs of RAM. 384MBs seems to work fine and 320MBs might be OK as well. Some of the current DGromacs work units that use SSE2 boost are an exception and can make better PPD than the "big packets" work units while using much less RAM when running on an SSE2 capable CPU.

Ragnar Dan wrote:

I've found a 128 MB stick is not large enough to consistently succeed with the SMP client, but it will probably work with the others, though I've not verified that. I simply know that when running notfred's SMP client, I often get write errors because the 128 MB stick is full.

I have run into similar issues running off of an older USB 1.0 256MB stick. The stick can fill up and start giving write errors. The other issue I have noticed is that it can take some times in excess of 5 minutes to load the data off a USB 1.0 stick after a reboot, where as my 1 GB USB 2.0 stick (Kingston 1GB USB 2.0, $9.95 + tax, free shipping) loads in about 15 seceonds.

http://www.buy.com/prod/kingston-1gb-da ... 43517.html

coffeecrazy wrote:

@notfred:

Thanks for the information about suitable USB stick sizes; I'll experiment with the 128MB stick, and invest in a few 1GB USB drives, just to cover the bases... BTW, notfred, do you happen to know the minimum amount of RAM required to make this work? I get the feeling that just about ANY size DIMM will do the job, but if you happen to know that a specific amount of RAM is required, I'd like to know what to budget for. I get the feeling I'm going to be doing some RAM experiments to see what effect using various sized DIMMs will have, and whether there is an improvement in work unit production by increasing the amount of RAM to a certain minimum threshold... What the hell, I should have some free time to research this, if no one else has had the opportunity yet...

Now, a question about some instructions from this webpage:

http://reilly.homeip.net/folding/usb.html

The instructions in question are:

"Edit the file isolinux.sys and change the USER and TEAM options to get it to fold for you rather than me. You may also want to change BIG=yes..."

I've downloaded the USB folding files (usb.zip and syslinux.exe) and examined them with 7-Zip; I didn't find any file named "isolinux.sys"... Is the actual file to be edited named syslinux.cfg? I opened it with a text editor, and (except for the different name) it seems to match up with the instructions; if those instructions do NOT apply to syslinux.cfg, can you point me in the right direction on locating a copy of isolinux.sys? If it is a text file, you can paste it here, and I can copy it into Notepad, and create a new version without having to download anything...

Other than that, it looks like I'm going to have to learn something about networking while I'm learning Linux... I've already downloaded and burned a few .ISOs (Puppy Linux and PCLinuxOS, just to name two) so I can start to get familiar with using Linux before I go back to New England...

Thanks again for all you've done to bring diskless folding to the clueless masses (well, ME, anyway), notfred; I'm looking forward to the day when I can FINALLY get my first diskless folder online... to be followed by others...

@Usacomp2k3:

Thanks for the welcome; I won't actually have a diskless system online until late spring at the earliest, but by learning everything I can now, I hope to be ready to start a diskless folding farm the day I get home. In the meantime, I've got a pretty steep learning curve ahead of me...

CoffeeCrazy - I find your story inspirational. What parts do you need to get your systems up and running? If I have any spare parts that can help your cause I will send them to you.

I am also trying to start up some USB folding on a 256mb drive, but when running the syslinux k:, the only file it creates before ending with an error is "ldlinux.sys". I've tried re-downloading, but that didn't help. Any thoughts?

Usage - syslinux K:

AppName: syslinux.exe AppVer: 0.0.0.0 ModName: syslinux.exe
ModVer: 0.0.0.0 Offset: 0000209c

It should write the ldlinux.sys file, but I'm not sure about that error, it may work anyway. You don't say what OS you are running on, but it looks like some variety of Windows.

It does create the file, all 11kb of it, and thats it. I am using XP Pro-32.

Edit: reformatted the usb to FAT32 from FAT, and it ran w/o errors.

I'll test the USB folding in a day or so...

ok, I set it up, was able to boot from the disk, then it seems to have gotten stuck in a loop detecting USB devices. Went through 3 iterations of devices 1-131, each time reporting a M$ wireless optical keyboard i have installed before I stopped it.

Proper usage is fold for a dual core, v.s. fold64 correct? As long as its set up properly in the config file that is...SMPCPUS=2 correct?

Jachyra007 wrote:

Proper usage is fold for a dual core, v.s. fold64 correct? As long as its set up properly in the config file that is...SMPCPUS=2 correct?

If you want to run the SMP client (preferably on real dual core CPUs not hyperthreaded fake ones ala the P4), there is only 64-bit available. Except for Core 1 Duo (Yonah), all the multi-core CPUs on the market today should be ok to do fold64.

Flying Fox wrote:

Jachyra007 wrote:

Proper usage is fold for a dual core, v.s. fold64 correct? As long as its set up properly in the config file that is...SMPCPUS=2 correct?

If you want to run the SMP client (preferably on real dual core CPUs not hyperthreaded fake ones ala the P4), there is only 64-bit available. Except for Core 1 Duo (Yonah), all the multi-core CPUs on the market today should be ok to do fold64.

Pentium D will also choke and die on x64

pikaporeon wrote:

Flying Fox wrote:

Jachyra007 wrote:

Proper usage is fold for a dual core, v.s. fold64 correct? As long as its set up properly in the config file that is...SMPCPUS=2 correct?

If you want to run the SMP client (preferably on real dual core CPUs not hyperthreaded fake ones ala the P4), there is only 64-bit available. Except for Core 1 Duo (Yonah), all the multi-core CPUs on the market today should be ok to do fold64.

Pentium D will also choke and die on x64

Really? They should have EM64T last I checked...

fold is for 32bit CPUs
fold64 is for 64bit CPUs
It will auto detect and run the correct one if you just leave it or hit enter without typing anything.

IF you are running fold64 AND you have more than 1 processor core THEN it will run the SMP client, ELSE it will run an instance of the normal client per processor core.

SMPCPUS is just a way of specifying how many instances of the SMP client to run on a machine with 4 or more processor cores. Divide the number of processor cores by SMPCPUS to get the number of instances of the SMP client that will be run.

That clears up the usage question - thanks!

How about the looping of USB device checking - is that normal? I only let it run for 3-5 minutes (3 loops) as I thought it was excessive. No screenshots avaliable, so if you need other clarification, I'll try my best...

Sorry, missed that. No the devices should all be detected within the first 10 seconds whilst it does "Waiting for any USB drives to start". If you see looping like that then there is a problem. Try unplugging any unneeded devices or only plugging the devices in once it gets to the waiting message.