Transforming RTLS Data Architecture to Big Data Architecture for Location Data

Real-time Location System (RTLS) locates and monitors any person or asset within a defined zone covered by a radio frequency (RF) network. This system produces location data for each entity to track from several times per second to every few minutes, depending on how often it is desired to refresh the location data. The increase in the volume of data that is required to be collected, stored, processed, analyzed, and reported, is one of the reasons for the development of big data technologies. This paper aims to explain the transformation of data architecture according to the target short-term and long-term analytical needs of RTLS and to compare traditional architecture and big data architecture on read times and storage costs. In order to achieve that, the data in the relational database system is extracted to.csv files in its raw form with a microservice triggered daily within the relational database management system. Data is processed in Python language to meet the three different reporting requirements of RTLS. Analyzed data is stored in a column-family NoSQL database Apache Cassandra by modeling differently in line with the query parameters for faster query purposes. By not archiving the location data on the relational database, without affecting the performance health of RTLS, meaningful results are produced from this data and are reported on the application. As the size of the input and output data used for reporting increases, the need to use a NoSQL database is essential in terms of query response times. NoSQL database systems may require the same data to reside in multiple tables, so the storage cost per row is about 67% lower than SQL database, but as the number of queries to the data increases, storage costs can exceed SQL databases.