An analytical model for delay bound of OpenFlow based SDN using network calculus

Abstract

Software Defined Networking paves the way for simplifying network management. This is achieved through the separation of control plane and data plane in a network. In Software Defined Networks (SDN), the network control functions are removed from the data forwarding nodes and placed into a logically centralized controller. So, these functions can be updated with a cost of a simple software change and without any changes in the hardware modules. But an important question arises about the performance of SDNs. This question becomes more serious regarding the central control unit as a probable obstacle to the scalability of SDNs. To have a precise image of this problem, in this paper an analytical model is provided for SDNs implemented based on pure OpenFlow standard. This model which is based on the network calculus framework, computes the worst case delay bound of such SDNs. The delay bound is estimated according to the similarities between the caches and flow tables in OpenFlow switches. The results show the interaction of different parameters such as network size, flow table size, traffic characteristics and the delay of SDNs. This can be used to evaluate SDN prototypes in the early stages of the design phase. Also, it affirms the importance of proposing scalable architectures for SDNs, such as deploying distributed controllers. Additionally, some points about using distributed controllers are discussed based on the provided results.