Benchmarking Materialized Views of SQL-based Stream Processing Systems

With the growing importance of real-time analytics, streaming processing systems are becoming an essential tool; in particular, the paradigm of the materialized view is gaining traction due to its simplicity and power. Moreover, a standard benchmark for these systems is still missing, although multiple benchmarks have been proposed for the more general field of stream processing. Many tools can support this kind of processing, both well-established and newer to the scene with different feature sets and technical choices. The goal of this thesis is to thoroughly evaluate and compare these tools in order to gain a deep understanding of their performances, characteristics, and limits. The competitors considered in this evaluation are Flink, Materialize, and ksqlDB (previously KSQL). Multiple aspects will be evaluated, from more abstract ones such as architecture, features, and technical choices to more concrete ones revolving around performance, scalability, and ease of use. This thesis proposes a baseline analysis of the three systems aim to exploit the advantages and disadvantages of each system and provide a comparison between them. As a starting point, we will use the Nexmark benchmark, which is a relatively simple and uncharted benchmark for stream processing systems.