Sensor-Rocks: A framework to improve software management for sensor networks

Software Operations and Management (O&M), i.e., defining, configuring, and updating thousands of software components within a conventional data center is a well-understood issue [12]. Existing frameworks like the Rocks toolkit [7], [11] have revolutionized the way system engineers define, deploy and manage large-scale compute clusters, storage servers, and visualization facilities. Unfortunately, the state-of-the-art approach in sensor network software O&M relies on system administrators to manage the operating system and overall system configuration of a system as a golden image. This image is manually configured according to site and individual project requirements. In the manual approach, this golden image is copied onto corresponding sensors typically using Over The Air Programming (OTAP) [8], [9], [10]. The fundamental problem with any golden image approach is that the methodology often used to build this software master is either unspecified or ad hoc. In either case, only the author of the golden image can change its contents, add new functionality, update existing components or completely rebuild (rebase) the image when the underlying OS changes. The manual approach does not scale, since ensuring consistency of installation, and accuracy of configuration and updates of thousands of software components is laborious and error-prone. One could dramatically reduce both time and effort while improving system reliability if techniques used to deploy scalable computing systems (e.g., data centers) were extended to the challenging world of sensor networks. We propose to develop Sensor-Rocks by adapting and extending the Rocks toolkit. We believe that Sensor-Rocks will revolutionize the world of sensor networks by reducing the administrative overhead of defining the software environment on individual sensors and network of sensors from days/weeks to a few hours. The resulting faster, reliable deployment of the software infrastructure will fundamentally improve the reproducibility, which is a key for gathering good-quality data.