A Distributed Wireless Control Plane for SDNs in Data Centers


Student thesis: Doctoral Thesis

View graph of relations


Related Research Unit(s)


Awarding Institution
Award date2 Oct 2019


Traditional data center networks require switches to perform control functions such as routing in a distributed manner. Software-defined networking (SDN) decouples the control plane from the data plane and centralizes the control plane via a controller. This allows each switch to only keep data plane functions such as packet switching. With SDN, network management has become easier and dynamic with the help of centralized and programmable control. To separate the traffic of control plane from the data plane, a separate physical network is required to connect all switches to the controller. However, it is an intricate task to build such a dedicated wired network for the control plane. It may incur enormous cabling effort to ensure the full connectivity between the all switches and the controller. To deal with this problem, we propose a wireless control plane design. In the proposed architecture, we install the ToR switches with wireless antennas and connect them to the SDN Controller via Access Points (APs) and Relay Nodes (RNs). We analyse the proposed wireless control plane architecture in three ways. (1) First we assume a semi wireless architecture where switches connect with APs using wireless channel(s) and APs are connected to the SDN Controller via cables. We analyse this semi wireless architecture by finding the analytical solution of cluster size. (2) Next, we see the effects of 802.11 Distributed Coordinated Function access schemes on the same architecture. (3) Finally, we find the analytical solutions for computing cluster size in a pure wireless control plane architecture.

In a semi wireless control plane architecture, switches are divided into clusters, and an AP is placed at the center of each cluster. An important issue is to determine the minimum number of APs such that a given control traffic demand can be met. We propose an analytical model to evaluate the system throughput for possible clusterings, and an efficient algorithm to search for the optimal one. The extensive simulations demonstrate that our method can reduce cabling complexity significantly.

In continuation to the work in semi wireless control plane design, we propose an analytical model to find the upper bound of cluster size. We analyze our model with respect to two different channel access modes of the IEEE 802.11 Distributed Coordinated Function (DCF) namely, Basic Access Mode and RTS-CTS Access Mode. We also analyze the effects of interference and channel assignment and obtain the lower bound of cluster size. Finally, we propose an algorithm to find the optimal cluster size that satisfy the control traffic demand of switches. With extensive simulations, we demonstrate that the proposed method can significantly reduce the cabling complexity.

In a pure wireless control plane architecture, the switches form clusters that are wirelessly connected to the controller via Access Points (APs) and Relay Nodes (RNs). The switches use 2.4/5 GHz band to connect with the APs, whereas the APs and the RNs are connected to controller using 60 GHz band. We have presented an analytical model to derive achievable data rates in our wireless control plane. We have also proposed two algorithms that allows an optimal number of cluster size of the switches to be connected with the Controller via minimum number to APs/RNs such that the control traffic demands of the switches is guaranteed in interference constrained environment. Through extensive simulations, the results of our propose architecture shows that the cabling complexity in the control plane is reduced to zero and additional switches may be easily added in SDN data center. Thus, a pure wireless solution for building a control plane in a data center network is feasible.