Efficient model checking of OpenFlow networks using SDPOR-DS

Abstract

OpenFlow is one of the most popular protocols to realize Software-Defined Networking. OpenFlow has attracted a great deal of interest because of its wide utility and applicability for automation of network management. While OpenFlow provides the ability to control a network using software, there is the risk of bugs occurring in the software that could cause erroneous network behavior. Therefore, improving the reliability of the network is very important. Model checking is a well-known technique to verify the correctness of distributed systems such as OpenFlow networks. However, it is difficult to apply it to this problem because model checking takes an exponential amount of time in relation to the scale of its target. Naïve model checking may take too much time, even to verify a toy network. We introduce an effective method for model checking of the OpenFlow network. Our method reduces the state-explosion problem with dynamic partial-order reduction and with state transition based on symbolic execution. We implemented a prototype for our method to evaluate it. The results indicated that our method completed model checking in less than 10% of the execution time of naïve depth first search model checking and in 31% of the execution time of an existing state-of-the-art tool.