BEE OS: Supporting Batch Execution with a Preemptive Real-Time Kernel

Abstract

Modern wireless sensor networks require operating system with real-time, low-power, multi-tasking features, in order to quickly deal with matters of wireless sensor nodes. Event-driven systems and multi-threaded systems are two major operating systems used by wireless sensor networks (WSN). To take advantage of salient features of both types, based on $\mu$C/OS-II and WSN characteristics, we have designed a tiny and hybrid embedded operating system, BEE OS, which integrates a preemptive real-time kernel supporting batch execution. Besides BEE has a real-time, low-power and multi-tasking feature. We propose a multi-task management scheme with multi-priority queues and analyze its timeliness performance based on a round-robin theoretical model, which leads to our novel design of the preemptive task scheduling algorithm to optimize the average waiting-time of all tasks. Finally, an adaptive method is introduced into BEE OS in order to avoid task starvation. Our experimental results show the overhead and performance comparisons among BEE OS, $\mu$C/OS-II, and TinyOS. Compared with $\mu$C/OS-II, the RAM and Flash of BEE OS are about two-fold down, and the task response time achieved by BEE OS is approximately a quarter of those by TinyOS and $\mu$C/OS-II. BEE operating system can be widely applied to relatively fixed tasks in industrial environments, with low power consumption and preemptive real-time characteristics.