2.1 Framework for analysis of the mechanisms and assumptions
A software solution containing unconnected logic components can be analyzed as acyclic graphs, disconnected and oriented, as mentioned in section Specific characteristics from a virtualization viewpoint. All logic components of the same solution must be part of a single acyclic graph, connected and oriented, whose vertices are the logic and virtual components, and whose edges represent dependencies between these components. The physical components in this study have no interest other than the fact that they are abstracted by virtual components. Section Criticality of a logic component explains the necessity that all the logic components be available to provide the totality of services offered by the software solution. Therefore, the framework for analysis will consider that the mechanisms will be evaluated for a software solution where all the logic components are essential to the solution, compared to virtual components, which can be duplicated at will and where each duplicate is not itself essential. The logic components of the solution will be processed without regard to their criticality vis-à-vis other components, the latter being abstract as mentioned in section Criticality of a logic component. Likewise, the analysis excludes the details of the internal processes of the logic components. This adds a characteristic to the connected and oriented graph: its vertices, representing the logic components, are not weighted. The availability analysis is done according to the premise of a total availability of all logic components, which are without internal error that can make them unavailable. Availability of 1 is then assumed for all the logic components, since there is a check on the internal implementation of the logic components, and the abstraction of any relationship with the external makes black boxes of these.
Contents · 10/26
- Design for High Availability
- Introduction
- 1 Hardware Abstraction
- 1.1 Isolating the levels of abstraction
- 1.2 Physical components category
- 1.3 Virtual Layer
- 1.4 Abstraction of virtual components
- 1.5 Characteristics of virtual components
- 2 High Availability Mecanisms
- 2.1 Framework for analysis of the mechanisms and assumptions
- 2.2 Public interface of a solution
- 2.3 Risk Control
- 2.4 One approach: duplication
- 2.5 Limits of replication of immutable virtual components
- 3 Evaluation of downtime risks
- 3.1 Dependency hierarchy
- 3.2 Bottlenecks
- 3.3 Method for calculating availability
- 3.4 Methodology of analysis
- 4 Evaluation in real cases
- 4.1 Basic software solution
- 4.2 Analysis without assumptions
- 4.3 Evaluation with addition of availability assumptions
- 4.4 Assumption in a cloud scenario
- Conclusion
- References