Spontaneous Investigation packet Creation

Abstract

Networks are getting larger and more complex, yet administrators rely on rudimentary basic tools to debug problems. We offer an automated and methodical tactic for investigationing and debugging networks, it is called “Spontaneous Investigation Package Creation” (SIPC). SIPC reads router configurations and creates a device-independent model. The model is used to create a least set of investigation packages to use every link in the network or exercise every rule in the network. Investigations are done occasionally, and detected failures trigger a separate mechanism to restrict the fault. SIPC can detect both functional like incorrect firewall rule and performance problems like congested queue. SIPC complements but goes beyond earlier work in static checking which cannot detect live ness or performance faults or fault localization which only restricts faults given live ness results). We describe our prototype SIPC implementation and results on two real-world data sets: AQ & T University’s backbone network and Internet2. We find that a small number of investigation packages suffice to investigation all rules in these networks: For instance, 4000 packages can cover all rules in AQ & T backbone network, while 54 are enough to cover all links. Sending 4000 investigation packages 10 times per second consume less than 1% of link capacity. SIPC code and the data sets are publicly available.