Mertens, Luca Alexander (2022) Reengineering Theodolite with the Java Operator SDK. (Bachelor thesis), Kiel University, Kiel, 63 pp.

Preview

Text bsc_luca_mertens_thesis.pdf - Published Version Download (1MB) | Preview

Abstract

Scalability is one of the most important quality characteristics of applications deployed in a distributed context. It is not trivial to predict how well a complicated system will scale under increasing load. Theodolite allows to efficiently benchmark the scalability of applications deployed on the container orchestration platform Kubernetes. Internally, Theodolite extends the functionality of Kubernetes by implementing the Kubernetes operator pattern. Since Theodolite started adopting operators, the pattern has matured, leading to the emergence of several best practices and tools. One of these tools, the Java Operator SDK, aims to simplify the process of writing Kubernetes operators in Java or Kotlin.
In this thesis, we apply the reengineering process to Theodolite’s operator: First, we analyze the state of the current operator and elicit its shortcomings. Second, we propose two architectures, one stateless and one stateful, for Theodolite’s operator and discuss their advantages and disadvantages. Third, we implement the stateful architectures using the Java Operator SDK. We evaluate our implementation by comparing the current operator to the reengineered operator in terms of performance and memory usage. We find that the reengineered operator brings both a performance improvement and a reduction in communication overhead with the Kubernetes API. We also measure a slight increase in memory usage in the reengineered operator.

Actions (login required)

View Item