Quote Originally Posted by D_A View Post
I'm afraid I don't think you can sneakernet WCG any more. Personally I'd just call it a "maintenance break" and use the time to clean heatsinks etc.
Personally I'd leave the machines online where they are if you can.
Sneakernet may still be possible, but it would be a PITA with OldChap's quantity of machines and data.
A quick search on "sneakernet" at WCG turned up http://www.worldcommunitygrid.org/fo...ffset,0#402557
The suggestion there would mean you'd take a snapshot of the BOINC data directory of each of your offline machines to the online machine. There you'd mount each snapshot and switch BOINC to use the foreign data dir & do its network stuff. Your offline machines would not be able to crunch using the directory with the data you've taken, until you return with the updated version. I guess you could have 2 data directories on each of your offline machines, and crunch into the alternate directory while the other one is away for transferring data.
You'd have to be very disciplined about which data directory is which. A PITA.
BTW, using BOINC's default 2 upload/download threads is very slow when there are lots of small files as for HCC GPU results, and it would take hours to upload your data. Increasing the no of upload threads should help: http://boinc.berkeley.edu/wiki/Client_configuration
<max_file_xfers>N</max_file_xfers>
Maximum number of simultaneous file transfers (default 8).
<max_file_xfers_per_project>N</max_file_xfers_per_project>
Maximum number of simultaneous file transfers per project (default 2).
HTH, but we're getting a bit OT. Maybe there should be a sneakernet thread.