SDN Survivability Research Based On Extended Open Vswitch

SDN (Software Define Networking) has become one of the important fields in the research of future work, and the core thought of it is to decouple network control and forwarding, optimize traditional network and even realize network reconfiguration. SDN survivability is one of the core problems in the developing SDN and the goal is to minimize the performance penalty as much as possible when the inevitable network failure happens.In the research of SDN survivability problem, there exist many challenges of long failure recovery time without control, network failure diversification, intra-domain and inter-domain survivability and info-updating when network failure happens. Based on the research and analysis of recent work and challenges, this paper uses OVS (Open vSwitch) as the basic architecture, which has been extensively applied in many SDN scenarios, such as SDN-based data center, and then we propose the SDN survivability architecture based on extended OVS. The main innovations in this paper are as following:1) This paper proposes the intra-domain SDN survivability mechanism based on extended OVS. To achieve this, this paper designs a connection state match and self-learning mechanism by extending OpenFlow and OVS. In connection state match mechanism, it can be aware of different connection states, achieve smooth handoff and process incoming packets by different scenario states. In self-learning mechanism,this paper extends it as OpenFlow action. It not only enhances packet-processing capability in the network data plane, reduces load of controllers and delay time, but also keeps the normal communication when the controller becomes unavailable. Therefore, the proposed mechanism improves the SDN survivability, and keeps the flexibility and philosophy of SDN.2) This paper proposes the inter-domain SDN survivability mechanism based on extended OVS. Based on the intra-domain stateful match mechanism, this paper designs an inter-domain forwarding mechanism and controller collaboration mechanism. By these mechanisms, this paper analyses two scenarios including in-of-domain and out-of-domain and resolves the inter-domain communication problems when a controller becomes not available.3) This paper proposes the info-updating reliability mechanism based on OVS after the failure recovery. This paper designs the info-dating mechanism for both controllers' database and switches. This can resolve the problem of stale information of topology and flow-tables'entries after the network recovers from failure.4) To evaluate the feasibility of our proposed SDN survivability mechanisms, this paper bases on OVS to make all of mechanisms modular and clearly describe how to implement them. To evaluate the performance of our design, this paper builds prototype and the experimental results show that the recovery time from controller connection loss is about 100ms level and it keeps low packets loss rate and a stable throughput. Therefore, compared with the traditional solutions, our proposed architecture enhances the survivability and stability of SDN.