Experimental Evaluation of Sketching Techniques for Big Spatial Data
Abu Bakker Siddique, Ahmed Eldawy · 2018
Ubiquitously connected devices, e.g., Internet of Things (IoT), space telescopes, social networks, and GPS-enabled gadgets, are contributing to the perpetual and swift growth of the data. 2.5 exabytes of daily-produced data, of which 60-80% is geo-referenced. Space telescopes broadcast about 140 GB of data weekly. Availability of such large amount of data calls for new scalable query processing techniques. One of the techniques that is getting attention is sketching which summarizes the data and computes an approximate answer on the sketch. This general technique is used in partitioning [3], clustering [1], selectivity estimation [2], and visualization [4], among others. This paper introduces a sketching-based framework for big spatial data which provides four sketching methods and uses them to implement three common operations, namely, partitioning, clustering, and selectivity estimation. The framework is executed in three phases, sketching, local operation, and generalization, which can apply to a wide range of operations on big spatial data.