initCluster: Initialize a Cluster

In gschnabel/clusterSSH: Launch R Code in Parallel on a Cluster via SSH

Description

Initialize a cluster running a job scheduling engine such as the Sun Grid Engine (SGE) software

Usage

initCluster(userFuns, remFun = NULL, sshOpts = NULL)

Arguments

userFuns	a list which contain at least the functions printId, start, and isRunning
remFun	object as returned from the initSSH function of the remoteFunctionSSH package
sshOpts	named list of arguments passed to remoteFunctionSSH

Value

Returns a list which contains the following functions:

start(fun, input, inputsPerTask=1, runOpts)

eval(fun, input, inputsPerTask=1, runOpts, pollTime)

result(job, selection=TRUE)

numberOfTasks(job)

closeCon()

status(job, raw=FALSE)

The meaning of the arguments is

fun	a function taking as argument a list input; it must return a list of the same length as input where each element of the resulting list contains the result associated with the respective element in input
input	a list where each element is an input for a calculation
inputsPerTask	how many elements in input should be submitted as one task
runOpts	a list with arguments for the SGE engine, which are cputime (seconds), vmem, and fsize (megabytes).
pollTime	pause in seconds between two consecutive queries of the job status
selection	select for which tasks the results should returned
raw	should raw output of qstat be returned

The return values of the functions are

start	returns job object
eval	returns the results associated with the elements of input
result	returns the same as eval
status	returns either a dataframe with information about the tasks or a character vector with the raw output of qstat

gschnabel/clusterSSH documentation built on Aug. 29, 2020, 7:01 p.m.