Patterns for Fault Tolerant Software

Chapter 4. Architectural Patterns

This chapter and the patterns it contains reflect architectural lessons from highly available systems of all sizes. The overall system architecture greatly influences how tolerant of faults the system will be. This chapter's techniques contain architectural considerations that cut across all parts of the system. They do not fit neatly into the categories of error detection, error recovery, error mitigation, and fault treatment. They do not focus solely on a particular class or module; they influence the design of the whole system.

These patterns are also among the first patterns applied to a new design project that needs to support fault tolerant operation. This is another reason for placing them here, in a separate chapter, rather that sorting them into the chapters of specific techniques.

For example, the pattern Minimize human intervention (5), which is found in this chapter, acknowledges that in order for a system to be highly available it must recover as quickly as possible. People are slow when compared with a computer. Asking a person to intervene in an error situation slows down recovery, which in turn reduces availability. A highly available system should not need to ask for human intervention. To reduce recovery times, the design of every part of the system must include the capabilities for the system to resolve the problems by itself. This need for a global usage of the principles results in the pattern being architectural. Every part of ...