Memory limitations

R is memory bound. This literally means that the dataset must fit into the client's RAM. From the previous example, if someone is going to run a simple algebraic operation, some matrix computation, or any kind of distance-based algorithm (that is heavy on computation), the R system would be limited to what would fit into the RAM. No spill to disk or any kind of temporary object is available, unless the user loads a package that allows otherwise.

Again, Microsoft has a solution with the RevoScaleR package that covers this limitation. With the introduction of the eXternal Data Frame (XDF) format, the package is able to store any size of the dataset (there is no recorder upper limit) in this format that is persisted on the disk, with conversion from a binary, blob, CSV, or any other format to XDF format. This limitation is made possible in Microsoft Machine Learning R Server or Machine Learning R Services.