NonStop Systems Introduction

The NonStop Kernel
NonStop Systems Introduction—527825-001
6-14
Process Pairs
messages to its backup in Processor 1. If Processor 0 fails, the primary disk process
fails. Processor 1 then informs the backup disk process that Processor 0 has failed,
and the file system automatically reroutes requests to this disk process, which has
become the new primary process.
As a result, the users of any programs that access the disk drive attached to the failed
processor (Processor 0) can still perform data retrieval and update operations on disk
files without any interruption. This fault tolerance would not be possible without the
reliable delivery of checkpoint messages by the message system.
Most operating system processes, like the disk process in Figure 6-9, are implemented
as process pairs, because all the application programs running in the system require
the continuous availability of these system processes for successful execution.
It is also possible to implement user processes as process pairs, and you can do so
without writing special code for the checkpointing.
Figure 6-9. Software Fault Tolerance With Process Pairs
Dual ServerNet fabrics
Processor 0 Processor 1 Processor 2
adapter
Primary
disk process
Backup
disk process
Checkpoint
messages
to backup
ServerNetServerNet
VST080.vsd
adapter