Performance Evaluation of Containerized Virtualization Platforms

Performance Evaluation of Containerized Virtualization Platforms

연구 분야: Software Development

학회: SN Computer Science

Cloud computing revolutionized how companies managed their applications by enabling access to computational resources over the internet, providing elasticity, scalability, and cost efficiency. For a long time, virtualization primarily referred to hypervisor-based methods. However, in recent years, container-based virtualization emerged as a widely adopted technique in modern software development, particularly within cloud computing. Tools such as Docker, Podman, and LXC/LXD enabled users to create, manage, and run containers. This study proposes benchmarking tests to analyze the impact of different container image sizes and assess the effects of using a hypervisor alongside containers. The research focused on how container management performance varied with different image configurations, exploring images of various applications in distinct sizes, and investigating correlations in performance across different OS-level virtualization tools. The findings highlighted the performance strengths of Docker, Podman, and LXC/LXD in different scenarios: Podman performed best with smaller images, Docker excelled in container stop and remove actions, and LXC proved higher start, stop, and remove times. These trends were observed in both physical and virtualized environments, with only slight performance differences in virtual machines.