PermaSense Data Management Jan Beutel, Mustafa Yuecel, Roman Lim, Tonio Gsell, ETH Zurich.

PermaSense Data Management

Jan Beutel, Mustafa Yuecel, Roman Lim, Tonio Gsell, ETH

Zurich

What we have today:

PermaSense Starting Points

Low-power Wireless Sensors

• Static, low-rate sensing (2 min)

• Temperatures, crack meters, resistivity

• 3 years operation• < 0.1 Mbyte/node/day

Base Station for Data Collection

High Resolution Imaging – Mountainview Camera

• Dual network (TinyNode, WLAN)

• 12 Megapixel D-SLR imager • Calibrated optics• Remotely configurable• ~ 1 Gbyte/camera/day

Long-haul Communication

• 7.5 km from weather radar on Klein Matterhorn

• Leased fiber/DSL from Zermatt Bergbahnen AG

• Collaboration with APUNCH/CCES

Data Backend Integration and Tools

• Based on GSN (EPFL research project)• Collaboration with SwissEx/EPFL• Slowly gaining in-house knowledge• Limited visualization capabilities• Many different other components necessary for 24/7 operation and debugging

The Big Picture – Network System Integration

Data Management Tools and Dataflow

Data Management – Online Semantic Data

• Global Sensor Network (GSN)– Data streaming framework from EPFL– Organized in “virtual sensors”, i.e. data types/semantics– Hierarchies and concatenation of virtual sensors enable on-line

processing– Translates data from machine representation to SI values– Adds metadata

Private

PublicMetadata==============PositionSensor type…

Import from field GSN GSN Web export

Multi-site, Multi-station Data Integration

TinyOS Multiplexing Data Flow

Example: Sensor Network and Backlog/CoreStation

Example: Private GSN Data Intake

Example: Public GSN Data Mapping and Conversion

Sensor Type Mapping

Sensor Node DAQ Routines

Simple - One Sensor Type - One Usage - One Packet

Harder - Two Packets, Multiple Sensor Types

Complex - Multiple Configs, Reuse of MUX resource

Position/Type Mapping: XYZ_nodepositions.xls

Open Problems

Open Problems - Random Order

• Merging of multi-packet data (ADCMUX1/2)• Multi-sink deployments

– Upstream/downstream data (protocol and control issue)• Multi-core station integration (proposal)• Visualization• Integration of further processing steps• Documentation• …