Agent-based workbench for on-the-fly sensor-data analysis

The emergent popularity of wireless sensor networks and cloud-computing services brings us new opportunities of integrating them for performing on-the-fly sensor-data analysis with as many computing resources as needed. Yet, users are required to integrate sensor networks and cloud services manually, and even to redesign their data-analyzing software for its parallelization. To alleviate these problems, we are developing an agent-based workbench that consists of three independent but interlinked software tools: (1) the AgentTeamwork-Lite job coordinator, (2) the MASS middleware library, and (3) the Connector elastic data channel. AgentTeamwork-Lite has each computing node exchange resource information with neighbors, and deploys a job with a mobile agent that searches for idle nodes. MASS allows programmers to view their applications as a series of interaction among autonomic computation entities that are dynamically mapped to a cluster of multi-core machines. Connector separates descriptions of network address, communication protocol, and data-sampling conditions from applications. It automatically streams and hands off sensor data to migrating jobs. This paper presents an overview of our workbench and its applicability to orchard business.