Extensible virtual machine for reprogramming wireless sensor networks

Abstract

Currently, wireless sensor networks are used for monitoring the environments that humans are difficult to access. Hence, changes in the environments have a dramatic effect on the data returned from sensor nodes. According to environmental changes, wireless sensor networks should be frequently reprogrammed. The characteristics of sensor nodes in wireless sensor networks are to be resource-constrained. Above all, power is one of the critical resources. So, reprogramming must be efficient from a energy point of view. We propose The Extensible Virtual Machine (EVM) to reprogram deployed sensor nodes with energy efficiency. The remarkable features of the EVM is dynamic modules for application specific processing logics. Modules reduce the overhead of interpretation , operand stack and TinyOS task queues in comparison with bytecodes for processing logics. The modules is dynamically linked to monolithic programm image. application developers can reprogram sensor nodes with modules as well as bytecodes. Our evaluation shows that modules achieve the performance improvement of about 20 for complex processing logics such as greedy forwarding algorithm and about 4 to 10 for simple processing logic such as avg., min., and max over bytecodes for same processing logics. Though modules have more update cost than bytecodes, it can be retrieved after several times of executions. Therefore, EVM provides energy-efficient reprogramming mechanism in spite of update cost.