Availability Guide for Application Design

Availability Guide for Application Design—525637-004
7-1
7
Availability Through Process-Pairs
and Monitors
Most applications do not need to use process-pair primitives directly. A combination of
the following products and features of an HP system provide the availability needed for
most classes of applications:
•
The supported transaction monitor—NonStop Tuxedo or the Pathway transaction-
processing environment—as described in Section 5, Increasing the Availability of
Tuxedo Applications, and Section 6, Availability in the Pathway Transaction-
Processing Environment
•
HP’s transaction-management tool, TMF, as described in Section 4, Data
Protection and Recovery
•
System hardware and software fault tolerance as described in Section 4, Data
Protection and Recovery
•
A process monitor as described in Using a Process Monitor on page 7-22
Almost all applications that use the tools described in Sections 4 through 6 make
indirect use of process pairs. In many cases, the ability to withstand transient errors
and other failures is provided by a process pair built into the tool you are using. Thus,
the complexity of designing and writing process pairs is hidden from the application
developer.
Although the transaction-monitor-based tool sets are extensive and flexible, they might
not be appropriate for all classes of applications. In these rare cases, you still have the
option of directly accessing the process pair primitives.
This section provides an overview of process pairs in application design and some
guidelines about when it might be appropriate to use them. Two models are described:
•
Passive backup, in which the backup process passively waits to be activated in the
event that the primary process fails
•
Active backup, in which the backup process actively updates its state while
listening for system messages
Most process pairs involve copying state information from the primary process to the
backup process. Because the mechanisms for doing this are significantly different,
different terminology is used to describe each mechanism. For passive backup process
pair, we talk about checkpointing information to the backup. For active backup
process pairs, we talk about updating state information in the backup.
After reading this section, you should have a better understanding of whether process
pair techniques would be suitable for your application design, which backup model
would be appropriate, and what kind of effort is required for different designs.