摘要
State-based models provide an attractive and simple approach to performance modeling. Unfortunately,this approach gives rise to two fundamental problems: 1) capturing the input loads to a system efficiently within such presentations; and 2) coping with the explosion in the number of states whtn system is compositionally presented. Both problems can be regarded as searching for some optimal representative state model with a minimal const. In this paper a probabilistic feedback search approach (popularly referred to as a genetic algorithm) was presented for locating good models with low (state) cost.
State-based models provide an attractive and simple approach to performance modeling. Unfortunately, this approach gives rise to two fundamental problems: 1) capturing the input loads to a system efficiently within such presentations; and 2) coping with the explosion in the number of states whtn system is compositionally presented. Both problems can be regarded as searching for some optimal representative state model with a minimal const. In this paper a probabilistic feedback search approach (popularly referred to as a genetic algorithm) was presented for locating good models with low (state) cost.
