G@H w/ no Internet Connection

[BaLa]

I want to run G@H # school on a couple Lab Comps
they're POSs but better than nothing..

we have a firewall I can not get through...

what files do I need to copy to move back and forth with me, so that I can return the workunits at home>?

[master7]

I dunno if it is the same for G@H, but for Folding@Home, all the work is in the /work directory

[JugHead]

copied from Genome@home Updates


Friday, November 2, 2001

One of the advantages of the Genome@home distributed computing project is that the client does not need to run on a machine with a constant internet connection. In fact, it can be run on machines with no internet connection at all. It wasn't a consideration in the original design of the client, but this capability was discovered by users early on in the development of the project, and further client development introduced some features which make the process of "nonetting" or "sneakernetting" slightly more convenient.

Since this has become a very popular and viable way to run the Genome@home client, I'd like to offer some instructions on how this can be done efficiently, without loss of processed work units. The protocol below assumes a setup consisting of one machine with an internet connection, the "net machine" and several machines without, "dummy machines". Any deviation from this protocol, or "variations on the theme", should be used with caution and careful forethought. Please note that, in general, this is a somewhat advanced technique, and you should only try nonetting if the instructions below make sense to you. There are thousands of possible variants of this, and we really can't effectively troubleshoot individual attempts or procedures.

Install the 0.99 client on ALL your machines. This is the most stable version, and contains the best features for nonetting. Download and install the client on the net machine. Run it long enough to go through the configuration step and let it get a cpuid from our server, and download a work unit. Once the rotamer filtering step has started, kill the client with Ctrl-C. Move the entire G@h directory to a dummy machine. It's fine to compress the directory when moving it.

Repeat step 1 in its entirety for all the dummy machines. This applies even to separate instances of the client on multi-processor machines. Each instance/processor should be treated as a separate dummy in this protocol.

Run the client on each dummy using "ghclient[.x] -nonet". This will allow the client to run and rerun indefinitely without attempting to connect with a server.

To "harvest" finished results from the dummies (I'd suggest about once a week), do the following. Move the set of completed results files to the net machine's G@h directory. These files are:

output.chi.######
str.######
csum.output.######


Be sure to erase these files from the dummy machine after transferring to the net machine. Again, compressing the files here is fine. Also copy the dummy id.dat file to the net machine's G@h directory. Run "ghclient -clear" on the dummy machine and Ctrl-C when it gives a "GetWork" error.

Upload the results from the net machine using "ghclient[.x]". Allow the client to download a new work unit, and kill the client with Ctrl-C when the rotamer filtering step begins.

Transfer the new work unit to the dummy machine by moving the following new files from the net machine to the dummy G@h directory:

input.pdb
input.inp
str.dat
seed.dat


Repeat from step 4 to harvest from each dummy machine.

Repeat step 3 to begin crunching the new units on the dummy machines.

Those familiar with nonetting will see that the order of operations in some of these steps can vary slightly. This is not the only method that will work, but it will work in most cases.

Happy Genoming!

:)

[BaLa]

screw all that...
I'll just copy the whole folder over once a week...
w/my ZIP drive...

added a P2-266 w/80mb RAM
LOL...
it works, it will help somewhat.. :D