File System Level Compression of Radio Space Information Storage System for Sensor Platform

Abstract

Rich participatory sensing applications by smart phones are demonstrating the possibility of useful applications with numerous stationary sensors as well as with smart phones. Electricity consumption of stationary sensors seriously affects their usability and maintenance cost so that many mutually incompatible wireless devices and protocols have been developed for each those different conditions. It is desirable for devices with any different protocol to share the network infrastructure, preserve sensing data, and jointly utilize the data. We proposed an "Appliance-defined ubiquitous network"' (ADUN) that, based on user demands, can distribute sampled RF data streams over the Internet to software-defined radio receivers in cloud data centers. One of the goals of ADUN is to allow users to be able to seek information regarding the radio space of any bandwidth, frequency, place, time, and date. An RF recorder is necessary to distribute past RF data, and should be able to record as broad an RF data stream for as long as needed. In this paper, we detail the basic concepts of RF recorder for ADUN and the results of a study that applies the Btrfs function in Linux to compress and store RF data to distribute or mine an RF signal through time-shifting. The experimental results indicate that the pipeline parallelism of Linux increases the storage writing throughput of high-bitrate RF data streams with some degree of redundancy, though the loss in computation power for RF data compression slows down the storage writing. The RF data compression rate is calculated by the size of the RF data, the chunk size in chunking, and variance in the radio space information according to the number of signals to be received.