Energy efficient sensor data logging with amnesic flash storage
Suman Nath · 2009
Flash storage based local data logging is beneficial for many sensing systems. However, flash size and battery capacity often limit the amount of sensor data that may be logged on a sensor node or an embedded gateway. An amnesic storage system can address this limitation by compressing and aging sensor data since sensor data is typically highly compressible and for most applications, older data is less valuable than newer data. While algorithms for compression and aging are well known, implementing them on flash leads to new challenges due to unique write and erase characteristics of flash memory. We show that existing amnesic compression schemes, although optimize for space or decompression error, are not suitable for energy-constrained devices. We present FlashLogger, an energy-efficient sensor data logging system that uses lazy amnesic compression in a flash-efficient manner. FlashLogger incorporates a suite of compression algorithms suitable for progressively compressing time series scalar, audio, and image data. All our methods are designed for the limited memory and processing capabilities typical of low power sensor nodes, and are prototyped on Tmote Sky platform running TinyOS. Evaluation of FlashLogger with several real world data sets shows orders of magnitude energy savings for both logging data and retrieving data within a time range. 1.