makeRserveCluster {Rserve.cluster}R Documentation

Create workers in a Rserve cluster by connecting to it.

Description

makeRserveCluster is creates a cluster to be used with parallel/snow which consists of usually multiple connections to machines running Rserve.

Usage

makeRserveCluster(names, ...)

Arguments

names

Either a character vector of host names of machines running Rserve to which to connect, or a positive integer (in which case that number of connections is made to "localhost"). If port option is set to 0 then names will be interpreted as unix socket paths (where supported). Mutliple listings are possible (and often desirable to achieve multi-core processing).

...

additional cluster options to set.

Details

makeRserveCluster creates a cluster which is recognized by the parallel. It connects to Rserve instances on the given hosts and all cluster operations essentially send asynchronous eval commands to the connected instances.

Rserve instances have a huge benefit that they incur no startup cost and that both code and data and be pre-loaded.

Value

Returns and object of the class RserveCluster which is essentially a list of RserveNode objects.

Note

Rserve instances must have the Rserve.cluster package at their disposal. It is also most efficient if the package is already loaded in the master instance (eg.e via "eval library(Rserve.cluster)" in the configuration file).

Author(s)

Simon Urbanek


[Package Rserve.cluster version 0.1-0 Index]