SensorDB is a collection of programs for easily accessing and working
with sensor data collected from Buildings.
Software Dependancies:
Python 2.6 (http://python.org/)
Matplotlib (http://matplotlib.sourceforge.net/)
Scientific Python (http://www.scipy.org/)
NumPy (http://numpy.scipy.org/)
Optional Software Dependancies:
LibSVM (for classification)
FastCluster (an optimized clustering library. If not
available will fallback on SciPy's clustering libraries)
Additionally, I would also very highly recommend the IPython shell
(http://ipython.scipy.org/) to which makes interactively playing with
large data sets a breeze.
As you might have noticed the libraries are written in
Python. Currently there are 2 data sets:
a. SCADA Sensor data collected from Soda Hall
for the period from Nov 2008 -- May 2011.
b. SMap Sensor feeds (http://smap.cs.berkeley.edu/) collected
from various sensors deployed across Cory Hall and the CITRIS
building.
The library has 2 data adapters TSDBTrace and FileTrace. The TSDBTrace
adapter connects directly to the LoCal Sensor DB and retrieves data
over the network. The FileTrace adapter on the other hand is optimized
to retrieve data lazily from trace files on disk.
The TSDB database is hosted on cluster running Hadoop's HDFS and uses
the super awesome HBase columnar storage and the OpenTSDB middleware
to store the sensor time series data.
Contributors:
Prashanth Mohan (http://www.cs.berkeley.edu/~prmohan)
David Culler (http://www.cs.berkeley.edu/~culler)