Failover Timing Analysis in Orchestrating Container-based Critical Applications

Emerging industrial applications require highly reconfigurable and flexible environments to effectively respond to dynamic market demands while complying with rigorous non-functional requirements. The orchestration of virtualized industrial components over edge/fog computing infrastructures allows the required reconfigurability to be achieved by enforcing automated management of hardware resources and applications. However, existing orchestration systems fall short of meeting crucial non-functional requirements, such as determinism, reliability, and application/criticality awareness, preventing their use in industrial environments. In this short paper, we conduct a preliminary timing analysis on failover mechanisms used in Kubernetes in order to identify sources of nondeterminism, which are fundamental to be mitigated in an industrial mixed-criticality scenario.