The sat command provides a demonstration of system performance and
capabilities. It is run as an application program with the operating system
booted for normal, multiuser operation. It is fully automated and runs on
all standard Paragon configurations.

The sat command may be run from an interactive shell or from shell scripts.
It does not request input from the user while running. Arguments to the
command select particular tests, the number of iterations, and other options.

By default, tests execute in the service partition and the entire compute
partition.  All processors in the system may be tested. With the system
otherwise idle, the tests run in a predictable, repeatable sequence
producing deterministic results.

Tests are stored in a directory hierarchy rooted by default at /usr/lib/sat.
This hierarchy contains subdirectories for each test with a well-defined
structure for sources, makefiles and run scripts. Tests also may be contained
in other directories that conform to the same structure, allowing users to
extend the acceptance test. The structure is documented in the manual.

Tests are self-configuring and run with appropriate parameters for the system
being tested. Non-parallel tests are executed in the service partition.
Parallel tests are executed on the largest set of nodes they will work on
in the compute partition. Test results include pass/fail status, performance
statistics, and any error messages.

A transcript is written during operation, allowing a user to track progress
and determine when errors occur. The transcript may optionally be written
to a log file. After completion, a detailed report of test results and
operating parameters may optionally be written to an output file.
