Proactive techniques for correct and predictable Internet routing

The Internet is composed of thousands of autonomous, competing networks that exchange reachability information using an interdomain routing protocol. Network operators must continually reconfigure the routing protocols to realize various economic and performance goals. Unfortunately, there is no systematic way to predict how the configuration will affect the behavior of the routing protocol or to determine whether the routing protocol will operate correctly at all. This dissertation develops techniques to reason about the dynamic behavior of Internet routing, based on static analysis of the router configurations, before the protocol ever runs on a live network. Interdomain routing offers each independent network tremendous flexibility in configuring the routing protocols to accomplish various economic and performance tasks. Routing configurations are complex, and writing them is similar to writing a distributed program; the (unavoidable) consequence of configuration complexity is the potential for incorrect and unpredictable behavior. These mistakes and unintended interactions lead to routing faults, which disrupt end-to-end connectivity. Network operators writing configurations make mistakes; they may also specify policies that interact in unexpected ways with policies in other networks. To avoid disrupting network connectivity and degrading performance, operators would benefit from being able to determine the effects of configuration changes before deploying them on a live network; unfortunately, the status quo provides them no opportunity to do so. This dissertation develops the techniques to achieve this goal of proactively ensuring correct and predictable Internet routing. The first challenge in guaranteeing correct and predictable behavior from a routing protocol is defining a specification for correct behavior. We identify three important aspects of correctness—path visibility, route validity, and safety—and develop proactive techniques for guaranteeing that these properties hold. Path visibility states that the protocol disseminates information about paths in the topology; route validity says that this information actually corresponds to those paths; safety says that the protocol ultimately converges to a stable outcome, implying that routing updates actually correspond to topological changes. Armed with this correctness specification, we tackle the second challenge: analyzing routing protocol configurations that may be distributed across hundreds of routers. We develop techniques to check whether a routing protocol satisfies the correctness specification within a single independently operated network. We find that much of the specification can be checked with static configuration analysis alone. We present examples of real-world routing faults and propose a systematic framework to classify, detect, correct, and prevent them. We describe the design and implementation of rcc (“router configuration checker”), a tool that uses static configuration analysis to enable network operators to debug configurations before deploying them in an operational network. We have used rcc to detect faults