xgrid: Submit a Set of Jobs to an Apple Xgrid
In xgrid: Access Apple Xgrid using R

Description Usage Arguments Details Value Note Author(s) References See Also

Submit a set of jobs to an Apple Xgrid, letting the grid deal with scheduling and load balancing.

xgrid(grid = "localhost", numsim = 20, ntask = 1, 
  indir = "input", outdir = "output", param = 1, 
  Rcmd = "runjob.R", auth = "None", outfile = "RESULTS.rds", 
  prefix = "RESULT", throttle = 9999, sleeptime = 5, verbose = FALSE)

`grid`	domain name of the grid controller
`numsim`	number of simulations (individual tasks) to run
`ntask`	number of tasks to be run by each job
`indir`	input directory (contents copied from the client to the agent by the controller)
`outdir`	output directory to be used to store results returned from the agent by the controller
`param`	an optional parameter to be provided to the remote job
`Rcmd`	R script file within the `input` directory to be run by the remote agent
`auth`	authorization to use to access the grid: options are "None" (default), "Kerberos" or "Password" (if "Password" then the environment variable XGRID_CONTROLLER_PASSWORD should contain the password)
`outfile`	directory to return output created by the remote job
`prefix`	file prefix to add to the beginning of the name of individual job results files
`throttle`	maximum number of jobs to run at one time (default = 9999), to avoid being greedy on a shared Xgrid
`sleeptime`	number of seconds to sleep after checking for completion of a remote job (default = 1 second)
`verbose`	verbose output (default = FALSE)

The function is used to generate a series of commands to execute using the Apple command line program xgrid which stuff jobs on a grid, monitor their progress, and return results when completed. The function assumes that the function Rcmd exists in the input directory, and that the return value for this function is stored in the object res0 that will be returned separately for each job and collated into a single dataframe.

The number of jobs run simultaneously can be controlled using the throttle option (otherwise, all jobs are queued and the Xgrid controller determines which are run).

Returns a dataframe with number of rows equal to the number of simulations (tasks) calculated as given by the numsim option. The specific contents of the dataframe are specified by the code run by the user.

It is possible to utilize add-on packages on the remote agent, even if no direct access to the machine is provided. This requires that packages are installed into the input directory, then accessed with the lib.loc option to library by the remote job.

Larger values of sleeptime may lead to longer job latency upon completion, but will provide less of a load on the grid.

Nicholas Horton (nhorton@smith.edu)

Mac OS X Server: Xgrid Administration and High Performance Computing Version 10.6 Snow Leopard (2009). Apple Inc.

xgridsubmit, xgridattr, xgridresults and xgriddelete

xgrid documentation built on May 2, 2019, 4:55 p.m.