Abstract: Duc to the mechanism that component based dataflow software has its execution path determined by components activated by data fed into input interfaces,its software reliability can be influenced by distribution of input data,and is thus difficult to evaluate using traditional models such as statcbased or path-wise ones. A reliability model combining frectuency statistics and operational profile was put forward in this paper. Starting from analysis of dataflow software structure, programs were denoted as multi-layer structure by introduction of composite node. On the basis of data and control flow relations,activated components and their execution probability were estimated. The operational profile was transferred along the direction of data flow to components on the same or a lower layer. By means of depth-first rule, reliability of composite nodes on all levels was evaluated in opposite order until the actual reliability of the whole software was obtained. A case study shows that the model can effectively evaluate actual reliability of component based data flow software, reflecting ready state of effective data in input interfaces and data distribution attributes.

Key words: Software reliability,Architecture-based reliability model,Dataflow software,Operational profile