NonStop Systems Introduction for H-Series RVUs

Integrity NonStop NS-Series Server Architecture
NonStop Systems Introduction for H-Series RVUs—540083-001
7-6
Interconnected Multiple Processors
The system achieves fault tolerance for the user’s database by ensuring that data
transfers between processes and storage devices can take place over two separate
and independent paths. In this way, if a component in one of the paths fails, the system
can use the other path to access the database. As a result, the failure of a single
component does not prevent access to the database, just as the failure of a single
processor does not halt the execution of processes.
The specific mechanisms used by the system to provide two separate data paths
include dual-ported ServerNet adapters (either on the ServerNet side or on the
peripheral side) and mirrored disks. As Figure 7-4 shows, mirrored disks are disks
that contain identical copies of the database. Although both disks are always in use for
optimum efficiency, if one of these disks fails, the process can continue accessing only
the other disk. For more information on mirrored disks, see Mirrored Disk Volumes on
page 7-11.
The following discussions take a closer look at the structure of the ServerNet fabrics
and the ways that they ensure the continuous availability of processes and databases.
Interconnected Multiple Processors
A NonStop system consists of 2 to 16 processors. Figure 7-5 on page 7-7 shows a
simple four-processor configuration. (For now, I/O is omitted.) Each processor in this
system is an autonomous computer in its own right, each having its own memory and
operating system software. The ServerNet fabrics make it possible for all the
processors in the system to communicate with each other and to cooperate in
executing applications. In this architecture, each processor is a ServerNet node and
each has a ServerNet node ID.