Quantitative Analysis of Design Decisions for Dynamic Reconfiguration

Dynamic reconfiguration is becoming an important requirement of current software systems as they have a trend towards running continuously and updating frequently. The main purpose is to reduce the update cost caused by system shutdown or restart. Controlling the influence of dynamic reconfiguration on system performance is an essential yet difficult issue. Many factors have to be considered and various methods could be chosen to use, including underlying component model, state transfer, connector type, reconfiguration algorithm and reconfiguration execution. These issues are analyzed in this paper and special attention is focused on how these design decisions affect the influence of dynamic reconfiguration on system performance. A DSE system is used as a case to analyze the design choices for dynamic reconfiguration through out this paper. A comparison of the influences resulted from different design choices is made through recording the influence parameters of the system in reconfiguration under different design choices.