Software that coordinates jobs or programs running on different nodes across the array.

High-performance parallel interface.

Interprocess communication. IPC provides sockets interface for low-level access for network and data transfers.

Load-sharing facility. A networking computing system running on top of heterogeneous UNIX systems that transparently schedules applications to improve performance and resource availability.

Message-passing interface. MPI allows processes to communicate by "mailing" data messages to other processes running on distant computers. Message-passing programs can execute on tightly coupled shared-memory multiprocessors, networked multicomputers, and wide-area distributed systems. MPI is a relatively new standard and is intended more for parallel computers made by the same vendor.

These are the individual systems such as the POWER CHALLENGE or POWER Onyx rackmount or deskside that form the fundamental component of the array. An array node is also known as a POWERnode.

Parallel virtual machine. A software package that permits a heterogeneous collection of UNIX computers connected by a network to be used as a single large parallel computer.

Various executing processes sharing memory on a virtual level.

The basic building block for sockets-based IPC. A socket is an endpoint of communication to which a name can be bound. Sockets are typed according to their communication properties, for example, stream sockets, datagram sockets, and raw sockets.

Symmetric multiprocessor.

Transport layer interface. TLI provides generalized functions that support network communications using many standard protocols. TLI is the AT&T version of Berkeley sockets. It interfaces with transport services and applications to provide client-server environments

X file server.