Container-Based Microservice Placement Optimization in Cloud
Name
Volodymyr Chernetskyi
Abstract
Today, many applications operate as containerized microservices running on multimachine clusters in cloud computing environments. High container-machine placement quality increases throughput, reduces latency, and improves cluster resource utilization. Optimal placement is achieved by solving complex multidimensional optimization problems. However, in response to the highly volatile nature of the cloud environment, modern container management systems opt for suboptimal placement, prioritizing quick decision-making. Over time, these poor decisions compound, increasing the need for placement reevaluation. Available reevaluation solutions change the placement of containers by making them undergo the same placement process again, with no guarantee of improved results. To address the issue, this thesis presents a novel placement reevaluation algorithm based on the particle swarm optimization approach with a focus on optimizing infrastructure costs. By integrating interzonal network traffic costs, the proposed algorithm achieves cost reductions beyond those obtained through conventional solutions that focus solely on computing resource costs. The stability of the algorithm in high-dimensional tightly bounded discrete search spaces was enhanced through improved particle position initialization. Empirical evaluations demonstrate the efficiency of the proposed algorithm, surpassing traditional optimization problem solvers, while outperforming standard particle swarm optimization implementations in terms of accuracy.
Graduation Thesis language
English
Graduation Thesis type
Master - Software Engineering
Supervisor(s)
Chinmaya Kumar Dehury
Defence year
2023