Processes within HP NonStop

This post explains the basic concept about a process on the Tandem. It also contains a (told) custom process presentation (Aiellogram) by Dennis Aiello, a tech trainer for Tandem computers from my training on the HP premises. The main conceptional focus is how to keep a process fault tolerant.

In computing, a process is an instance of a computer program that is being executed. It contains the program code and its current activity. A critical operating process, e.g. disk process, run within the HP NonStop (previously Tandem) as process pair spanned over two CPUs:

  • The primary process (1 CPU)
  • The backup process (1 CPU)

Any process that encounters a fault, e.g. the CPU crashed, is automatically switched to the backup instance. When the backup has taken over, it will become the primary, and creates a new backup instance, in case the new primary may also crashed. This can go on until every CPU is exhausted. In theory this might happen, but in reality, having e.g. 8 real CPUs crash simultaneously or after each other is unlikely. The process on a Tandem has following characteristics:

  • A process name (if none given, automatically assigned by the OS) like $data1, the backup process has the same process name!
  • A process identifier in ,
</em> notation like 1,714 (cpu 1, pin 714) * A priority (Range 1-255, User Range: 1-199, System Range: 200-255), Order by highest number (desc) * A code space and data space (see diagram) * A process owner by notation group.user. The process user super.super is like root, the manager of the system. In numbers it is in this notation (255,255) * A terminal from which the process was started