- Author: David de Andres Hernandez, deandres.hernandez@tum.de
- About: Paper submission for the seminar course 'Communication Networks from LKN, TUM
When cloudifying any system, the choice of an appropriate container network solution is critical. The wrong container network choice can turn a working system unviable and eclipse the main benefits of cloud architectures: reliability, scalability, and flexibility. Yet the number of solutions targeting the cloud environment is vast and the angles approaching the challenges are varied. Therefore, understanding and evaluating the benefits and limitations of the available solutions and technologies is a tedious but crucial process. To support this process, this paper surveys nine published benchmarks and three performance models searching for the major affecting factors and estimators. The findings are the following: the CPU prevails as bottleneck for performance in most scenarios; although many performance benchmarks exist, not many target scenarios with large-scale container pairs; performance models for containers with focus on the CPU resources provide the most precise estimators.