Because early design decisions can have a major long term impacton the performance of a system, early evaluation of the high-levelarchitecture can be an important risk mitigation technique. This paperproposes a technique for predicting the volume of data that will flowacross a network in a distributed system. The prediction is based uponanticipated execution of scenarios and can be applied at an extremelyearly stage of the design. It is driven by requirements specificationsand captures dynamic metrics by defining typical usage patterns in termsof scenarios. Scenarios are then mapped to architectural components, anddataflow across inter-partition links is estimated. The feasibility ofthe approach is demonstrated through an experiment in which predictedmetrics are compared to runtime measurements
展开▼