Localized Explanations for Automatically Synthesized Network Configurations

Network synthesis simplifies network management by automatically generating distributed configurations that fulfill high-level intents. However, typical network synthesizers operate as monolithic algorithms, obscuring the internal workings of the synthesis process and showing no clear connection between the generated configurations and the global intents. Given the critical role of networks as infrastructure, it is crucial for network operators to understand the synthesized configurations to establish trust in these automatic tools. To address this challenge, we propose using subspecifications localized to each component in the network topology to enhance the interpretability of network synthesis. These subspecifications provide insights into the workings of synthesizers by connecting each component's functionalities with the global configuration intents.