Availability analysis of blade server systems
IBM Systems Journal
Performability analysis of storage systems in practice: methodology and tools
ISAS'06 Proceedings of the Third international conference on Service Availability
| Hi-index | 0.01 |
As the new generation high-availability commercialcomputer systems incorporate deferred repair servicestrategies, steady-state availability metrics may no longerreflect reality. Transient solution of availability modelsfor such systems to calculate interval availability overshorter time horizon is desirable. While many solutionmethods for transient analysis have been proposed, howto apply these methods on hierarchical models has notbeen well addressed. This paper describes an approachto computing interval availability and related metrics forhierarchical Markov models. The approach divides thetime interval of interest into small subintervals such thatthe input parameters can be treated as constants in eachsubinterval to make the model satisfy the homogeneousMarkov property, and then pass the output intervalavailability metrics as constants from the sub-model to itsparent model. Finally, these quantities are integrated toobtain the expected interval availability for the entireinterval. The study also addresses methods of passingparameters across levels for generating multiple metricsfrom a hierarchical model. The approach is illustratedwith an example model and has been implemented inRAScad. All computations for the example model havealso been carried out using the SHARPE textuallanguage interface.