sleigh {nws}R Documentation

Class “sleigh"

Description

Represents a collection of R processes used to execute tasks.

The sleigh allows R functions to be executed in parallel using the eachElem and eachWorker methods.

The sleigh workers are started when the sleigh object is constructed. When tasks are submitted to the sleigh, using the eachWorker and eachElem methods, the workers will execute the tasks, and return the results. When the stopSleigh method is called, the workers are stopped.

Note that a given R program can create multiple sleigh objects, which will each have it's own set of workers. This could be useful if tasks have different requirements. For example, you could create a Linux sleigh and a Windows sleigh, and Windows-specific tasks only to your Windows sleigh.

Objects from the Class

Objects can be created by calls of the form sleigh(...), where `...' can be one or more of the following named arguments:

  • nodeList: a list of hosts that workers will be created. Note that this argument is ignored when launch equals `local'. The default value is c('localhost', 'localhost', 'localhost').
  • workerCount: number of workers that will be created. This argument is only relevant when launch equals `local'. The default value is 3.
  • launch: method to launch remote workers. This can be set to the string 'local', 'web', or to a function object. The function will be called once for each worker listed in nodeList. It will be passed two arguments: a name from nodeList, and an environment object that was constructed by merging defaultSleighOptions with the arguments that were passed to the sleigh constructor. The function should return a character mode vector, where the first element is the command to execute, and the subsequent elements are the command arguments. For example, the function could return the vector: c('ssh', '-f', host), where 'host' is the first argument to the launch function. This isn't the complete command that will be executed: it's the 'remote execution' portion of the command. The sleigh constructor will add rest of the command based on the scriptExec argument. Note that the command is expected to return after launching the worker. That is why the ssh -f option is used in the example.

    The default value of launch is `local'.

  • nwsHost: host name of the netWorkSpaces server. Default is the machine where sleigh starts up.
  • nwsPort: port number of the netWorkSpaces server. Default is 8765.
  • scriptExec: command to execute worker script. Default uses scriptcmd function on Windows, and uses envcmd function on other platforms.
  • scriptDir: location of the sleigh worker script. Default is the bin directory under where nws library is installed on the system. If library cannot be found, then use current working directory.
  • scriptName: worker script file name. Default is RNWSSleighWorker.py on Windows, and RNWSSleighWorker.sh on other platforms.
  • workingDir: worker's working directory. Default to master's current working directory.
  • logDir: location where log files will be stored. Default is NULL.
  • outfile: remote workers' standard errors will be redirected to this file. Default is NULL.
  • wsNameTemplate: template name to create sleigh workspace.
    Default is 'sleigh_ride_%010d'.
  • user: user name used for remote execution. Default is NULL.
  • verbose: a boolean value indicating whether to print out debug messages. Default is FALSE.

    Methods

    initialize
    signature(.Object = "sleigh"): sleigh class constructor.
    eachElem
    signature(.Object = "sleigh"): evaluate the given function with multiple argument sets using the workers in sleigh.
    eachWorker
    signature(.Object = "sleigh"): evaluate the given function exactly once for each worker in sleigh.
    rankCount
    signature(.Object = "sleigh"): get sleigh's rankCount.
    status
    signature(.Object = "sleigh"): return the status of the sleigh.
    stopSleigh
    signature(.Object = "sleigh"): shutdown workers and remove sleigh workspace.
    workerCount
    signature(.Object = "sleigh"): get number of workers started in sleigh.

    Details

    There are five different standard launch types ('local', sshcmd, rshcmd, lsfcmd, and 'web') to tailor client's working environment. This is done by setting launch variable to a function (sshcmd, rshcmd, or lsfcmd) or a string ('local' and 'web'). See the examples section.

    Examples

    ## Not run: 
    # Default option: create three sleigh workers on local host:
    s <- sleigh()
    # which is equivalent to:
    s <- sleigh(launch='local')
    
    # Create sleigh workers on multiple machines using SSH:
    s <- sleigh(c('n1', 'n2', 'n3'), launch=sshcmd)
    
    # Use the LSF bsub command to launch ten workers:
    s <- sleigh(launch=lsfcmd, rep('fake', 10))
    
    # Use web launch:
    s <- sleigh(launch='web')
    ## End(Not run)
    

    [Package nws version 1.6.3 Index]