A new quantitative approach to the problem of reconfiguring a degradable multimodule system is presented. The approach is concerned with both assigning some modules for computation and arranging others for reliability. Conventionally, a fault-tolerant system performs reconfiguration only upon a subsystem failure. Since there exists an inherent trade-off between the computation capacity and fault tolerance of a multimodule computing system, the conventional approach is a passive action and does not yield a configuration that provides an optimal compromise for the trade-off. By using the expected total reward as the optimal criterion, the need and existence of an active reconfiguration strategy, in which the system reconfigures itself on the basis of not only the occurrence of a failure but also the progression of the mission, are shown.Following the problem formulation, some important properties of an optimal reconfiguration strategy, which specify (i) the times at which the system should undergo reconfiguration and (ii) the configurations to which the system should change, are investigated. Then, the optimal reconfiguration problem is converted to integer nonlinear knapsack and fractional programming problems. The algorithms for solving these problems and a demonstrative example are given. Extensions of the optimal reconfiguration problem are also discussed.

/pdf/optimal-reconfiguration-strategy-for-a-degradable-3fydjawc60.pdf

Optimal reconfiguration strategy for a degradable multimodule computing system

A new method to determin an optimal retry policy and for use in retry of fault characterization is presented. An optimal retry policy for a given fault characteristic, which determines the maximum allowable retry durations to minimize the total task completion time was derived. The combined fault characterization and retry decision, in which the characteristics of fault are estimated simultaneously with the determination of the optimal retry policy were carried out. Two solution approaches were developed, one based on the point estimation and the other on the Bayes sequential decision. The maximum likelihood estimators are used for the first approach, and the backward induction for testing hypotheses in the second approach. Numerical examples in which all the durations associated with faults have monotone hazard functions, e.g., exponential, Weibull and gamma distributions are presented. These are standard distributions commonly used for modeling analysis and faults.

/pdf/optimal-design-and-use-of-retry-in-fault-tolerant-real-time-25phj764je.pdf

Kang G. Shin

Papers

Optimal design and use of retry in fault tolerant real-time computer systems