Static Differential Program Analysis for Software-Defined Networks

Networks are increasingly controlled by software, and bad updates can bring down an entire network. Network operators therefore need tools to determine the impact of changes. To address this, we present static differential analysis of software-defined network (SDN) controller programs. Given two versions of a controller program our tool, Chimp, builds atop Alloy to produce a set of concrete scenarios where the programs differ in their behavior. Chimp thus enables network developers to exploit the power of formal methods tools without having to be trained in formal logic or property elicitation. Furthermore, we show that there are many interesting properties that one can state about the changes themselves. Our evaluation shows that Chimp is fast, returning scenarios in under a second on several real applications.