Reconfiguration for Sensitivity Technique (RST) is presented as an approach for maintaining QoS in stream-based applications running on embedded systems. It is a co-design approach that uses the reconfiguration and sensitivity metrics for QoS awareness; the former for run-time reconfiguration in response to stimuli from the latter. The effectiveness of RST was tested after we used our CHARMS algorithm to screen out 95.17% of the mapping-cases for a H.263 encoder. Our tests showed a 99.25% QoS provisioning-up-time-level at any given instant, based on available battery power, using RST as compared to 80.62% with Performance Aware Reconfiguration of Software Systems(PARSY), 58.83% with Reconfigurable Service Composition(RSC) and 41.67% without RST.