Basic mass deployment

From BOINC

For a computer to run BOINC, it must have

  • The BOINC client program (boinc.exe on Win, boinc on Unix).
  • A "data directory" in which the client runs.
  • In this directory, configuration files that connect the client to accounts on one or more BOINC projects, or to an account manager.

To run BOINC on a number of "worker" computers, create these files and directories on each computer. If the computers share a network file system, they can use the same copy of the client program, but each computer must have its own data directory.

Configuration files

To create the configuration files:

  • If you haven't already done so, create accounts on the projects you want to compute for, or on the account manager.
  • Run the BOINC client on a desktop machine (Win, Mac, or Linux).
  • Attach it to those projects, or to the account manager.
  • Locate the BOINC data directory (e.g. C:/Program Data/BOINC on Win); the configuration files will be in that directory.

If you're attaching to projects, each one will have a file with a name like account_www.gpugrid.net.xml.

If you're using an account manager, you'll need two files: acct_mgr_login.xml and acct_mgr_url.xml.

If you want to use non-default computing preferences on the worker computers, set them on the desktop machine, and include the file global_prefs_override.xml.

Copy only these files to the worker computers.

Running BOINC on the worker computers

If you want BOINC to run all the time on each worker computer, you'll need to arrange for the client to run at system startup; how to do that depends on the OS.

If you're running BOINC as low-priority "backfill" on a cluster, you may want the client to run a single job per CPU, then exit. To do so, run the client with the --fetch_minimal_work and --exit_when_idle command-line options.