articles

< Back to articles list

The author: Namiot D. Published in № 2(68) 29 april 2017 year
Rubric: Software development technologies

Buy an article

Download the first page

Time series databases for Internet of Things

This article discusses issues related to the databases for time series. Time series are widely used in various applications. Accordingly, the theme of time series persistence is very important. The description of time series persistence and processing from the perspective of the database is the subject of this article. This paper deals with the use of traditional relational databases to store time series, as well as solutions that are classified as NoSQL systems. Of course, under certain conditions, which are discussed in this article, we can use traditional relational databases. Possible limitations are primarily associated with the new data accumulation rate. In this connection, we discuss fractal trees as optimal indexes for time series in relational databases. As the useful data processing tool, we describe widow functions and other SQL extensions for time series processing. On the other hand, possible limitations and peculiarities of processing time series lead to the need to use specialized systems for storing time series. NoSQL solutions present a rich set of tools for time series processing. We discuss data models (structured files, column-based databases) as well as data formats and network protocols for time series. As a general result of our survey, it can be concluded that the choice of solutions for time series databases is determined, primarily, by the rate of arrival of new data. From NoSQL solutions, Cassandra is the preferred choice.