Konkel, Christopher (2023) Benchmarking Scalability of Load Generator Tools. (Bachelor thesis), Kiel University, Kiel, 55 pp.

Preview

Text bsc_christopher_konkel_thesis.pdf - Published Version Download (686kB) | Preview

Abstract

Scalability is one of the most important quality characteristics of modern day applications. It is a non-trivial effort to predict performance changes and scaling capabilities of complex systems under increasing load. As a solution, a wide variety of load generator tools can be used to simulate high amounts of load on a system under test. But in a testing environment with limited hardware resources, the system under test is not the only component of interest in terms of system resource requirements, as load generator tools can also require significant amounts of hardware resources. Theodolite, a framework for benchmarking horizontal and vertical scalability of cloud-native applications, also uses a load generator for its benchmarking process.
In this thesis, we compare the vertical scalability of the custom load generator used by Theodolite to the three load generator tools k6, Gatling and JMeter using Theodolite benchmarking methodology. We evaluate the tools based on the niche use case of the custom load generator, simulating input of collected sensory data. We evaluate the HTTP and Kafka load format. To achieve this, we first configure the three tools to match the selected use case in both load formats and then integrate them into the Theodolite environment. Second, we define new benchmarks for Theodolite to use the selected load generators in its benchmarking process. Third, we execute the newly defined benchmarks and evaluate the results of the experiments. We find significant differences regarding the vertical scalability of the selected tools, with scalability differences of the same tool depending on the load format.

Actions (login required)

View Item