Optimal Deployment of Cloud-native Applications with Fault-Tolerance and Time-Critical End-to-End Constraints

Cloud environments are becoming increasingly interesting to host time-critical use cases with far more stringent latency requirements than conventional cloud-native applications, such as smart industrial control systems or cloud-enabled autonomous vehicles. In these emerging domains, fault tolerance mechanisms play a critical role, due to the catastrophic consequences a fault might lead to, in the real world. This work presents a formal model for designing and deploying time-critical, cloud-native applications under fault conditions. Our model considers the interactions and interferences among service components and the possible occurrence of faults. We present an optimization framework to solve the deployment problem of minimizing the resources needed to achieve fault-tolerance under precise end-to-end deadline constraints. The ability of the optimizer to deliver precise temporal and fault-tolerance guarantees is validated through extensive simulations.