Various kinds of dynamic routing architectures are used in today’s service- and cloud-based architectures, including sidecar-based routing, routing through a central entity such as an event store, or archi- tectures with multiple dynamic routers. We propose an analytical model of request loss during router and service crashes, as well as an empirical validation of that model. The comparison of the empirical data to the predicted values by our model shows a low enough and converging error rate for using the model during system architecting. Our model predicts that, having the same crash probability, decentralized routing results in losing a higher number of requests in comparison to more centralized approaches. To the best of our knowledge, our study is the first to em- pirically study the reliability trade-off in such architectural decisions.

Top