Adaptive parsing of router configuration languages

Abstract

Network functionality is growing increasingly complex, making network configuration management a steadily growing challenge. Router configurations capture and reflect all levels of network operation, and it is highly challenging to manage the detailed configurations of the potentially huge number of routers that run a network. One source of difficulty is the constant evolution of router configuration languages. For some languages, particularly Ciscopsilas IOS command language, and its relatives, these changes demand frequent maintenance of configuration parsers in any configuration management tool. The essential problem is that config parsers understand a statically determined set of inputs, requiring human intervention to modify that set. We propose an alternative design for router configuration parsers: adaptive parsers. Such parsers can infer the configuration language based on real configs and automatically adapt to changes in the config language, all with minimal human involvement. We present the design of such a parser and discuss its prototype implementation for the Cisco IOS configuration language. We have validated our prototypepsilas accuracy and efficiency by running it on the configuration files of Tier-1 ISP networks. Our results show that from only 81 configuration files, we can learn enough IOS to successfully parse all of the 819 IOS configurations in under 10 minutes.