Today we live in a world, where our everyday activities are becoming more and more digitalised. We are witnesses of an age, where we have a large number of devices at our disposal, which are connecting to the web and are producing large quantities of data. Among these devices belong not only laptops and mobile phones, but also even more present sensors in cars, smart homes, industry and power plants, which are sending rows of data into the web. Traditional databases could not cope with all the challenges, which were presented by such large amounts of data. Thus the timeseries databases emerged. Their main characteristic is that they add a timestamp to sensor data, which is then used as a basis for grouping. Finally this allows them to handle data more effectively. Currently, there are many different timeseries databases and others which support work with timeseries. Consequently, this greets the developers with the difficulty of choosing the right database for some solution. Therefore to fulfill this task, the developers require the assistance of a benchmark. A benchmark is basically a program, which through a defined and controlled environment, evaluates the performances of some software, with a set of tests. In the case of databases, this represents a program which prepares data and executes a set of given queries, on a database. Lastly it measures the effectiveness, through the chosen criteria. The problem is that, there are merely a few existing benchmarks, which could be used as a reference, as the TSDBs are a relatively new concept. These are usually limited to only a few databases and scenarios. The goal of this thesis is to present a benchmark, which will through a use case study, help answer the question which timeseries database is more suitable for a given scenario.
|
