Institutional Repository
Technical University of Crete
Technical University of Crete
EN
|
EL
| URI: | http://purl.tuc.gr/dl/dias/1C9D7FF5-D9E1-4ED4-8A15-C1AB774EE9C1 | ||
| Year | 2025 | ||
| Type of Item | Master Thesis | ||
| License |
|
||
| Bibliographic Citation | Konstantinos Tsakos, "Design and evaluation of microservice placement strategies in cloud infrastructures", Master Thesis, School of Electrical and Computer Engineering, Technical University of Crete, Chania, Greece, 2025 https://doi.org/10.26233/heallink.tuc.103073 | ||
| Appears in Collections |
The design and evaluation of microservice placement strategies in cloud infrastructures is crucial for optimizing resource utilization, reducing operational costs, and ensuring high performance. This thesis explores the Service Placement problem on cloud environments, with a focus on improving resource allocation and minimizing egress traffic. Service Placement is modeled as a graph clustering problem, and various clustering algorithms are investigated—specifically Affinity Propagation, Maximum Standard Deviation Reduction (MSDR), and Markov Clustering—and are combined also with a placement strategy called Heuristic Packing to develop efficient service placement solutions.The study is based on two benchmark microservice applications, iXen (IoT prototype) and Online Boutique (e-commerce platform), deployed on Kubernetes clusters in a Google Cloud environment. Through load stressing, criteria like the performance of different placement strategies in terms of node utilization, egress traffic reduction, execution time, and cost efficiency are evaluated. The results show that Affinity Propagation with Heuristic Packing and Maximum Standard Deviation Reduction with Heuristic Packing consistently outperform other strategies, offering low resource utilization, reduced egress traffic, and minimal costs in total.The findings suggest that while Affinity Propagation provides fast execution, making it suitable for dynamic environments, MSDR offers superior long-term optimization at the expense of execution time. These strategies are recommended for applications with high inter-service communication and varying traffic loads. This work contributes to the field by providing insights into the application of clustering algorithms for microservice placement and suggests future directions for integrating machine learning and adaptive strategies to further optimize service deployment in cloud-based systems.
Copyright © DIAS 2013
