Compute model: files

Files are inputs and outputs to applications, and the executable application programs. Each file has a name that is unique within the project. A file may also have one or more URLs, indicating where it is (or will be) located on data servers. Once a file is created (on the project server or on a participant host) it is immutable.

Files have the following Boolean attributes:

Some files originate on hosts; i.e. as output files of applications. If the Upload-when-present flag is set, the file (when available) is uploaded to one of its URLs. If the Sticky flag is set, the file remains on the host (where, e.g., it will serve as input to a later computation).

Other files originate in the back-end servers, and are placed on data servers. If the Sticky flag is set, the file is retained and a participant host after it is used.

The core agent and scheduling server exchange descriptions of files. These descriptions include an MD5 checksum of the file, providing protection against corruption and spoofing.