Six Dimensions Of Benchmarking: New Perpectives On Time Series Databases for Monitoring Applications
by
413
IPE
Masster thesis presentation of Sandro Melissano on the topic : Six Dimensions Of Benchmarking: New Perpectives On Time Series Databases for Monitoring Applications
To be able to assess the capability of any database system, we first need to understand how they work and then benchmark them with exactly the workloads they will deal with in production. In this presentation, I present my development of an improved benchmark with novel features with which I tested (benchmarked) four different database architectures, all specialized on time series data for monitoring applications. For the first time, a benchmark stressed systems comparing the workload performance of regular time series vs irregular ones, comparing them with data of various data-cardinalities (means: data points, which have multiple values each) and comparing them under a certain type of mixed workloads (writing, meanwhile reading and calculating aggregations). All of these workload characteristics are present, everyday, in application’s like KATRIN’s monitoring system.
Finally, we evaluate, which database system and which architecture is the most suitable one for this use case.
Lecture language: English