POSTER: Dynamic Dataflow Topology Monitoring for Real-time Seismic Networks

Lindquist, K.G., T.S. Hansen, B. Ludaescher, F.L. Vernon

Complex real-time seismic networks often develop an extensive set of dataflow connections, especially when one includes the virtual networks established by data-sharing. As these systems grow, the task of managing existing connections and adding new ones has the potential to rapidly overwhelm network operators and system architects. This is amplified by data-source deletions and additions and temporary network problems, to say nothing at all of the the changing presence of individual data streams.

We have created dynamic status-monitoring tools for Antelope-based seismic networks, expanding on capabilities of the Antelope system, which allow us to monitor the topology of orbserver connectivity and the dynamic dataflow for individual streams. These tools are important for monitoring network health, avoiding common traps such as data loops, and engineering sensible data-flow architectures.

Status monitors run on each orbserver, delivering metadata about server connectivity, clients, and data streams to a central site. These metadata are saved in a near-real-time relational database. We generate directed-graphs from this database. From these, using the GraphViz graphical processor, we create interactive overview images of connectivity and dataflow.

We present the current status of this technology; challenges we have encountered in its development; and application to multi-signal-domain sensor grids. We discuss the potential for further growth, including more sophisticated analysis algorithms and the possibility of automated, dynamic sensor-network reconfiguration.