The parallel-processing architecture provides high performance for database operations on computing platforms that range from a computer with a single CPU to parallel-processing platforms composed of dozens or hundreds of computers.
A parallel-processing platform is a set of independent computers that operate in parallel and communicate over a high-speed network, bus, or interconnect. The database server can run on several types of parallel-processing platforms, including:
An individual computer that operates within a parallel-processing platform is referred to as a node. A node can be a uniprocessor or an SMP computer. Within a computer of this type, resources can be segregated into smaller independently addressed subsystems with separate CPUs, memory regions, and I/O channels. Independent subsystems configured within an SMP computer are considered to be nodes for purposes of database server configuration.
You can configure the database server on a single computer or a parallel-processing platform as a set of one or more coservers. A coserver is the functional equivalent of a database server that operates on a single node. Each coserver performs database operations in parallel with the other coservers that make up the database server. Each coserver independently manages its own resources and activities such as logging, recovery, locking, and buffers.
The database server does not require that any hardware resources be shared between nodes. This approach is sometimes referred to as a shared-nothing architecture.
Figure 2 illustrates the database server parallel-processing architecture.