A novel method for parallel indexing of real time geospatial big data generated by IoT devices

Abstract

IoT produces a huge amount of big data as it comprises billions of devices that are interconnected with each other through internet. Today’s majority of the big data part is about geospatial data and every year it increases rapidly. In order to process such massive real time geospatial big data, we must have scalable, efficient indexing method. R Tree and its variants have emerged as most efficient, widely accepted and have adopted indexing method for the management and processing of geospatial data. Current literature on parallel construction of R Tree indexes of geospatial data has disadvantages, that all the methods considered only two dimensional geospatial data and all are based on MapReduce framework. As the number of dimension increases, complexity of index creation is also increases along with this MapReduce framework has lots of disadvantages such as, it works only on static data, consumes a lot of disk space and time, which leads to high latency and fault tolerance of the entire system. In order to overcome these issues, a novel method for parallel construction of R Tree and its variants, use of the Apache Spark (in-memory and on-disk computation) based on the IoT Zetta platform is proposed. The main purpose of using Apache Spark is to index real time geospatial data for continuously updating the position of aircraft in real time while indexing it in R tree and its variants, so that spatial range query can fetch real time results and Apache Spark is much faster as compared to MapReduce framework. The extensive experimental results show that our parallel generated R tree and its variants retains similar properties as of sequential generated R tree and its variants with the excellent scalability and reducing a significant amount of time for the construction of index, index updating & executing spatial range query over geospatial data by exploiting the latest parallelism framework.