Simplify Stochastic QoS Admission Test for Composite Services through Lower Bound Approximation

Abstract

A composite service can have its overall Quality of Service (QoS) measure computed with the QoS measures of its constituent services. In the stochastic case of QoS modeling, accurate computation for the probability distribution of the composite QoS measure is NP-hard because of the inherent complexities of probability value calculation for the function of discrete random variables. However, given reasonable assumptions on the monotony of the composite QoS function and on the independence of constituent QoS measures, we have proposed a lower bound approximation algorithm that computes the approximate value of the composite QoS distribution for admission test purpose in much lower-order complexity of time even in the worst case. The effectiveness of the proposed method is verified and compared against the naive algorithm using simulative trace data.