Admission control in software-defined datacenter network in view of flow table capacity
Chuanji Zhang, Hemin Yang, George F. Riley · 2018
Software-Defined Networking (SDN) is an emerging technology providing flexibility and programmability to network. It is envisioned as a key enabler for future network innovation, especially for datacenter network. However, the wimpy capacity at switches, such as flow table capacity, brings challenges to the scalability of OpenFlow-based SDN. The large gap between the current flow table capacity and the number of needed flow entries in a datacenter requires a novel admission control mechanism in view of the switches' flow table capacity, instead of only considering the link capacity as in traditional network. In this paper, we proposed a controller level admission control mechanism addressing this problem, and its effectiveness is evaluated based on the numerical simulation results using a canonical three-tier datacenter topology. The simulation results indicate that with limited flow entry and buffer capacity at the switch, the proposed mechanism can provide service to 42% more users and increase the total received data by 23-25%. This lightweight mechanism also improves the network goodput by reducing the number of packet loss at the switch.