Events
Andreas Terzis
Johns Hopkins University
Department of Computer Science
Baltimore, MD, USA
Andreas Terzis is an Associate Professor in the Department of Computer Science at Johns Hopkins University, where he heads the Hopkins InterNetworking Research (HiNRG) Group. His research interests are in the broad area of wireless sensor networks, including protocol design, system support, and data management. Dr. Terzis is a recipient of the NSF CAREER award.
Abstract
K2: A System for Campa ign Deployments of Wireless Sensor Networks
Environmental scientists frequently engage in “campaign- style” deployments, where they visit a location for a relatively short period of time (several weeks to months) and intensively collect measurements with a combination of manual and automatic methods. During 2009 we developed and deployed K2, a mote-based system that brings high-quality automated monitoring to deployments of this nature. A 50-node K2 network was deployed for 5 weeks starting in November of 2009 at the Nucleo Santa Virginia research station in the Atlantic coastal rain forest near São Paulo, Brazil. The results from this first deployment indicate that K2 is a viable scientific tool, achieving data yield > 99% and producing accurately time-stamped data, even in the absence of a persistently available reliable clock source. These results point a path towards WSN deployments managed by non-CS specialists.
Multiple sensor nodes, each made up of a TelosB mote, external antenna, battery, and sensor multiplexer board form the bottom tier of the K2 system. Nodes take ADC samples from up to four external sensors at a fixed frequency, compress them, and store them in local flash. Nodes periodically exchange local time references with each other, but otherwise keep their radios off to save energy when not participating in downloads. When the researchers’ schedules permit it, they bring the basestation laptop to the field site. The basestation wakes up the sensor nodes, builds a centralized view of the network topology, and downloads any new data from nodes it can reach over multi-hop source-routed paths. If an Internet connection is available, researchers upload the collected data to the back-end server. This server-class machine hosts an SQL database of the data collected thus far and performs the necessary translations from compressed data in motes’ local time scales to physical values in the global time scale.
The storage subsystem of K2 is tailored specifically to the requirements of campaign deployments. We would like to record sensor measurements with the highest possible fidelity (i.e., raw ADC measurements), but we also want to ensure that the data recovery rate is loosely coupled with the rate of site visits: we don’t want to lose data because the researcher couldn’t make it to the field for a day. To achieve these goals, K2 implements a data-centric delta compression component. The data is collected with a modified version of Koala. K2 differs from Koala in its use of a weighted (rather than thresholded) link selection scheme and random breadth-first download order (which favors “fresh” over “stale” link information). This approach supports disconnection-tolerance by quickly adapting to the basestation’s location as the researcher takes it to different locations in the network (e.g., if the network does not form a single connected component). In contrast to many collection protocols that assume a persistent basestation and routing tree, K2 nodes maintain a low duty-cycle when there is no basestation (one transmission every 20 seconds). We support data recovery by building reliable delivery on top of the unreliable data stream primitive offered by Koala: each download attempt consists of the primary download of buffered data followed by a data gap-recovery phase during which the basestation re-requests data that was not received during the first phase. The data retrieved from the motes is first collected into a “preliminary” dataset in the field and is later uploaded to a database and processed into a “science-ready” dataset. The preliminary dataset uses a single calibration curve for all sensors of the same type and uses only the basestation-to-mote time references to do timestamp reconstruction. This setup requires minimal configuration on the part of the field scientists, but still gives them enough information about the data being collected to adapt the deployment (e.g., by replacing or moving hardware). This promotes data recovery by identifying problems with data collection before the end of the deployment. The researchers upload the data from the basestation to a web application, which decompresses the data and inserts it into a database. At the end of the deployment, we use Phoenix to assign timestamps to the data and per-sensor calibration curves to convert measured values to physical values.
Looking into the future, the biggest challenge is scaling the system to deployments that are at least an order of magnitude larger than the one performed in 2009. Doing so will require innovations in all aspects of the system: drastically reducing the mote costs, scaling the network protocols to support larger deployments and dynamic re-programming and implementing self-management mechanisms.