Analysis of Virtualization Performance on Resource Efficiency Using Containers and Unikernel

Applications on a computer require resources and robust security measures to prevent attacks. In environments where multiple applications run simultaneously, isolation becomes essential, and the computer must efficiently manage resources for these isolated applications. Virtualization technology addresses this need by providing virtual resources, enabling the operation of virtual computers without requiring additional physical hardware. This approach enhances flexibility, efficiency, and secure isolation in managing computing resources. This research investigates the performance differences between container-based and unikernel-based approaches to virtualization, specifically on the Google Cloud Platform (GCP). The study is conducted in two stages: general tests and special tests. General tests involve measuring the image size and boot time, both critical factors in optimizing system performance and resource allocation. Special tests assess the system's performance under load, focusing on CPU and memory utilization by generating simulated requests to an Nginx web server using the wrk benchmarking tool, which simulates multiple concurrent clients to create a realistic workload. The results reveal significant advantages of the unikernel approach. The Nginx image size in a unikernel environment is about 1 MB, making it 57 times smaller than in a containerized environment. Additionally, unikernel boot time is 169 times faster. Furthermore, the unikernel demonstrates 22% more efficient CPU utilization and 0.6% better memory efficiency compared to containers, highlighting its potential for improved performance in specific use cases.