Reducing the Time of Live Container Migration in a Workflow

As a lightweight virtualization solution, container technology can provide resource limiting capabilities and can run multiple isolated process sets under a single kernel instance. Multi-tenant preemption leads to competition in computing, storage, and network resources, resulting in degraded computing service performance. Virtualization service migration can provide a solution to the problem of resource shortage in supercomputing systems. However, the resource overhead and delay in the migration process also reduce the efficiency of high-performance computers. To solve this problem, this dissertation proposes a method of container migration and a tool for supporting container migration. Also, this paper optimizes the startup of containers from checkpoints and proposes a multi-container migration strategy, reducing migration time by 30% compared to sequential migration. The migration method in this paper provides valuable experience for service migration in supercomputers and data centers.