DEFINED: Deterministic Execution for Interactive Control-Plane Debugging.

Large-scale networks are among the most complex software infrastructures in existence. Unfortunately, the extreme complexity of their basis, the control-plane software, leads to a rich variety of nondeterministic failure modes and anomalies. Research on debugging modern control-plane software has focused on designing comprehensive record and replay systems, but the large volumes of recordings often hinder the scalability of these designs. Here, we argue for a different approach. Namely, we take the position that deterministic network execution would vastly simplify the control-plane debugging process. This paper presents the design and implementation of DEFINED, a user-space substrate for interactive debugging that provides deterministic execution of networks in highly distributed and dynamic environments. We demonstrate our system's advantages by reproducing discovery of known ordering and timing bugs in popular software routing platforms, XORP and Quagga. Using Rocketfuel topologies and routing data from a Tier-1 backbone, we show DEFINED is practical and scalable for interactive fault diagnosis in large networks.