All complex system development projects involve analysis of the system architecture. Thus far it has been assumed that there is some correct system decomposition that can be used in the architectural analysis without consideration of the sensitivity of the results to the chosen level of decomposition. We represent 88 idealized system architectures and a real complex system as a design structure matrix at two different levels of decomposition. We analyze these architectures for their degree of modularity. We find that the degree of modularity can vary for the same system when the system is represented at the two different levels of granularity. For example, the printing system used in the case study is considered slightly integral at a higher level of decomposition and quite modular at a lower level of decomposition. We further find that even though the overall results can be different depending on the level of decomposition, the direction of change toward more modular or more integral can be calculated the same regardless of the level of decomposition. We conclude that the level of decomposition can distort the results of architectural analysis and care must be taken in defining the system decomposition for any analysis.
- system architecture