Abstract: This paper deals with embedded systems software and the modification of its architecture and behavior at execution-time. Incautious implementation of such capabilities demands heavy memory and performance overrun. To accomplish such software evolution activities in resource-constrained embedded systems, we propose a component-based run-time evolution infrastructure that reconciles richness of evolution alternatives and performance requirements. Our proposition is based on off-site components reifications, representations of components which allow us to treat evolution concerns remotely and hence to alleviate the workload to be processed by the embedded device. Memory and processor-time resources consumption evaluation on a real-world scenario show the efficiency and pertinence of our approach.