Performance Modeling of Microservices with Circuit Breakers using Stochastic Petri Nets

Circuit Breakers (CBs) are critical for containerized microservices infrastructures when managing a surge in workload, as they can serve as a protection mechanism to prevent system overload and meet service level agreements (SLAs). In the event of a sudden increase in workload, microservices (MSs) can face challenges such as resource exhaustion and many discarded requests. CBs can help mitigate these issues by monitoring the services and, if necessary, redirecting the flow of requests to another infrastructure. However, it is difficult to evaluate the behavior of these infrastructures and CB mechanisms in a production environment. To address the above issues, this paper proposes a model using stochastic Petri nets (SPNs) to represent these infrastructures and their auto-scaling mechanisms, the MSs, the CBs, the incoming external arrival rate, and the workload generated between the MSs. Service providers can estimate metrics including circuit breaker activation (CBA), overload forwarding rate (OFR), containers utilization (Uc), unallocated containers (NUc), throughput (TP), discard probability (DP), and discard rate (DR). The model enables the performance evaluation of individual MSs and the entire microservice platform (MP). The work investigates how the microservices adapt to changing conditions and the trade-offs associated with different CB configurations. Using a real testbed, our solution was validated with a confidence interval (CI) of 95%. A case study was used to investigate the feasibility of the solution by evaluating its application in a real-world scenario. We found that the CBs reduced DR by 71.4% on average. This corresponds to an average number of 75,454 requests over 1 hour that were not discarded but forwarded to another infrastructure.