Application Profiling for Automated VM to Container Migrations

The migration of applications from virtual machines (VMs) to container environments offers significant performance and scalability benefits but remains a challenging process. It often requires manual dependency collection—a time-consuming and error-prone process. Existing automation solutions, such as Google Migrate to Containers (M2C), follow a lift- and-shift strategy but often produce bloated containers by including unnecessary system-level dependencies, undermining the core benefits of containers.

This study introduces a novel process-level profiling methodology to automate the migration of applications from Linux-based VMs to containers. The approach combines static and dynamic analysis of application runtime behavior and focuses exclusively on the active Unix processes associated with the target application to identify critical dependencies, ensuring lean and optimized container configurations.

Case studies on NGINX and MySQL demonstrate the efficacy of the proposed approach, achieving container size reductions of up to 98% compared to Google M2C while maintaining or improving application performance. To validate its scalability and practical applicability, a command-line tool was developed based on this methodology. While effective, the method may face challenges with non-deterministic workloads, where certain runtime dependencies might be missed, so improvements are needed to ensure consistency. Nonetheless, this research represents a significant step toward streamlined, efficient, and accessible container migration.