SRGM建模类别与性能分析

Analysison SRGM modeling categories and performances

针对软件可靠性增长模型SRGM(software reliability growth model)在可靠性评估与保障中的重要作用,为全面掌握SRGM的建模与工作机理,对SRGM的典型建模过程以及不同模型间的性能差异进行深入研究.首先剖析了SRGM建模的基础假设和含义,梳理了SRGM的发展演化历程;然后分析了两类基本SRGM建模流程与关联,针对考虑更多真实测试情况的建模趋势,对不完美排错相关与考虑测试工作量TE(Testing-Effort)相关的SRGM建模过程进行了剖析;最后选取8个典型的模型在4个失效数据集上进行实验,依据度量与拟合结果进行了模型差异化的深入分析.研究分析表明,客观上不同失效数据集间的差异以及主观上研究人员对测试过程认知的差异是造成SRGM性能差异的主要根源.进一步建立更为准确与全面的SRGM,在有限的数据集上选取出优秀的SRGM已成为当前研究中亟待解决的难题.

In terms of the importance of SRGM （ Software Reliability Growth Model ） in evaluating and ensuring reliability, in order to grasp the modeling and working mechanism of SRGM, the typical process of SRGM modeling and the differences of performance in different models are studied in this article. First, the fundamental assumptions and the meaning of SRGM modeling are illustrated, and the development of SRGM is summarized. Second, the modeling processes and the relationship of two basic types of SRGM are analyzed. For the tendency of considering more real testing factors into SRGM, the SRGM modeling process relative to the imperfect debugging and TE （ Testing-Effort ） are discussed. Finally, the performances of 8 typical models selected are compared using 4 published failure data sets, and analyses on the differences are illustrated. The results indicate that the objective differences in different failure data sets and subjective differences in cognition of testing process by different researchers are the main causes that account for the different performances of SRGMs. Further establishing a more accurate and comprehensive SRGM and selecting excellent ones on finite failure data sets are the problems that must be solved in the future.