Hadoop-based distributed sensor node management system

We present a new architecture of HDSM (Hadoop-based distributed sensor node management system) for distributed sensor node management using Hadoop mapreduce framework and distributed file system (DFS). It offers various efficient ways for collecting sensor data and managing multiple sensor nodes by launching specific mapreduce applications on sensor nodes which upload data of sensor nodes to DFS and retrieve sensor data periodically from DFS. Additionally, it provides a flexible management scheme for sensor node by reconfiguring firmware or updating configurations and data formats of sensor nodes based on mapreduce framework. Also, it achieves a transparent and fault-tolerant architecture by exploiting various crucial features of Hadoop. Our experimental results show that it has efficient and stable performance.