A New Software Reliability Growth Model Predicated on Counting Processes for lnstruction Execution

概要

論文の詳細を見る
We have developed a new software reliability growth model based on counting processes for instruction execution in software. Our model defines the reliability of the program at time t, R(t), as R(t)=∏__<i⋴s>exp[-P_i/(a+p_i)⋋q_it] where S ={G, L, C, O}, G, L, C, O denotes global, Iocal, communication, and others, respectively ; a is the total number of faults ; p_i is the probability that failure occurs at the first execution of a class-i instruction, proportional to the number of faults in class-i instructions (we call q_i the initial probability of failure) ; λ is the rate of instruction executon ; and q_i is the probability that an instruction execution is of class-i. Through analysis using the proposed model, we conclude that higher software reliability can be achieved with data abstraction techniques than with functional decomposition, under reasonable assumptions. We note also that the exponential NHPP model is a special case of our theoretical model, and that the results of our model therefre agree with those of the NHPP model.
一般社団法人情報処理学会の論文
1994-12-15