This paper presents a novel architecture for native process migration (PM) in wireless sensor networks (WSNs) without the use of virtual execution environment. Resources in WSN are scarce; therefore creating virtual execution environment puts extra burden on already stringent resources. In addition, the proposed architecture is migrating with complete process instead of code only which also saves resources. The proposed architecture makes process migration decisions by continuously monitoring resources, such as remaining battery life and free memory space on a node. The architecture is suitable for networks with fewer expensive sensor nodes as it allows for better utilization of network resources. Transferring a live executing process from one node to another to meet processing demands dynamically improves fault tolerance, resource utilization, and network management in WSN. The architecture has been successfully tested and implemented on both COOJA simulator and a test bed of TelosB motes.
Platform-independent profiling in a virtual execution environment