The existing research work mainly focuses on pervasive cloud migration architectures and decision algorithms, but lacks the exploration of in-depth context awareness for pervasive cloud environment and comprehensive migration architecture.In this paper, we introduced a deep context awareness approach based on our previous work on OSGi-based elastic pervasive cloud infrastructure;Proposed a prediction method for available resources, using deep learing network of Long Short-Term Memory (LSTM) ;Designed a service migration mechanism on the basis of the monitoring and prediction of resources, in order to optimize services execution;Finally, tested and analyzed the migration of services in the cloud environment to show the effectiveness of our approach.