In-memory data is transported to the nodes along with he tasks, so it is immediately accessible without need for reading it from or writing it to a specified source.

URL-based data is accessed through a URL, where the URL specifies the location of the data, and the actual data is provided as an input stream. This method takes advantage of the JDK's built-in support for multiple protocols, including HTTP, FTP, Jar, File and Gopher.