Available for Windows only at the moment and for R version 3.0 — install as follows:
install.packages (“rinfinispanapi”, repos=c(“https://r-packages.coherentlogic.com/”))
Imagine being able to use The R Project for Statistical Computing with virtually no upper limit on the amount of memory you have to work with. If you find this idea attractive then keep reading.
JBoss, a division of Redhat, sponsors the Infinispan project — an open source implementation of shared-memory data grid technology that competes with proprietary (and typically expensive) platforms developed by companies such as Microsoft, IBM, Oracle, and other vendors. Infinispan is completely open source and is written in Java, has been in development for several years, and has many customers.
The R Infinispan API Package offers users of the R Project for Statistical Computing the ability to easily store incredible amounts of data. This package achieves this by tying into the Infinispan platform and offering R users functions which hide integration-specific complexity.
Some of the benefits this package delivers include:
- Store and access data in-memory with everyone in your network using simple keys and values.
- Expand the memory available to everyone simply by starting another instance of R, or by starting new Infinispan nodes anywhere on your local network.
- Seperate R scripts such that loading data and analyzing the data do not need to be tied together.
- Take advantage of existing infrastructure written in Java without having to tie directly into custom frameworks — if the data can be serialized to the grid, then all R users will be able to utilize it using the key it’s stored under.
- Fault tolerance is a configuration option — set the number of copies to two and two copies of data will be kept on the grid. Restart an instance of R and the data will still be available!
- Avoid continuous disk-access and queries that are required when working directly with a database. If the data is not available, get it from the database, add it to the grid using this package, and from that point forward anyone who needs it will find that it is already there for them to use.
If your network supports IP multicast then you can be up and running on more than one machine in less than five minutes.
Some examples of Infinispan running in R below.