Shulou(Shulou.com)06/01 Report--

The concept of SDN has been popular for many years, from the initial laboratory product to 2012 when Google announced that its backbone network was fully running on OpenFlow, increasing the utilization of wide area lines from 30% to nearly saturated. Thus it is proved that OpenFlow is no longer just a research model in academia, but has fully possessed the technology maturity that can be applied in the product environment.

The strength of the SDN community was still very weak in 2014, with only floodlight, Ryu and the very immature opendaylight. Scientific research institutions and universities are still using mininet to do simulation experiments. Later, with the carrier-led ONOS controller, Juniper OpenContrail and Huawei's DragonFlow, and the more mature OpenvSwitch, the situation has improved, and X86 servers can be used to build small SDN networks.

Up to now, there are many white-brand switch manufacturers, with many excellent commercial SDN controllers, and not only in data center scenarios, but also SD-WAN and other WAN scenarios, and now the concept of SDN is also expanding. With the popularity of OpenStack cloud computing, excellent open source projects such as Neutron components, and integrated Plugin with a large number of functions are also overturning the traditional definition of SDN. With the popularity of SDN and the landing of cloud computing projects, it has to be said that SDN is now used in the production environment of a large number of users. Follow the Kyushu Cloud Wechat official account "Kyushu Cloud 99CLOUD" for a white paper on SDN-related cases, and many Kyushu Cloud SDN practices are analyzed in detail.

Selection of SDN Network

The advantages of SDN network will not be discussed here. At present, SDN has many networking solutions to meet different user scenarios.

(1) the controller controls the forwarding of the white card switch through the southbound protocol.

The first implementation is that the implementation of the pure southbound protocol does not support any traditional protocols. Through the SDN controller to discover the whole network topology, calculate the path, and then send the flow table through the southbound protocol to realize the traffic forwarding according to the policy. This is shown in the following figure.

In the cloud computing scenario, the SDN controller can use OpenStack Neutron to implement the control plane, and control VXLAN VTEP, GW and IP GW through OpenFlow/Netconf. Its advantage is that it can realize the functions of network creation, subnet division, route selection and firewall policy management through softswitch. This is shown in the following figure.

(2) loose control mode scheme (self-learning through network equipment control protocol)

Automatic establishment of VxLAN tunnel and automatic tunnel association; tunnel establishment and address learning are completed through the standard E × × ×, neighbor discovery is realized with E × × × BGP RR, each device advertises its own VxLAN information, and each VTEP device has network-wide VXLAN information and the relationship between VxLAN and the next hop. The VTEP device automatically establishes an VXLAN tunnel with the next hop that has the same VXLAN as itself, and associates this VxLAN tunnel with these same VxLAN. The SDN controller is only responsible for issuing the service policy, not the control flow table, so it is more reliable. Note that VxLAN belongs to the Underlay for network transmission.

The MAC/IP route is transmitted to the peer VTEP via MP-BGP. In reality, the BGP connection is required to be full mesh (arbitrary pairwise interconnection), but in order to reduce the pressure of configuration, BGP RR (Router Reflector) is usually introduced. The role of BGP RR is to reflect the data of one BGP Speaker to all other connected BGP peer. Using BGP RR, you can make all BGP Speaker only need to establish a connection with BGP RR, otherwise according to full mesh, any BGP Speaker must establish a BGP connection with all other BGP peer.

(3) VLAN model networking

In the environment of OpenStack, Neutron network components can choose VLAN networking model. The characteristic and disadvantage of the VLAN model is that it needs to release the corresponding VLAN on the traditional network switch, so it needs to rely on the network planning when making the network access, at the same time, it is not particularly convenient in the multi-tenant isolation, so it still needs to be operated on the network switch, that is, the network networking is not flexible enough; the advantage is that the VLAN technology is relatively mature, whether in performance or stability.

The last words

In the data center, there are roughly three ways to deploy a software-defined network. One way is to use OpenFlow/Netconf, a southbound standard for SDN, but is often criticized for its poor scalability. A more popular way is to use virtual network coverage, even if VxLAN Overlay is used to expand networking. This technology is the most widely used at present. Another mainstream technology is based on an open protocol called BGP E, which isolates traffic between virtual machines belonging to different tenants in the data center. It is very powerful because it enables dedicated connections between virtual machines located in different parts of the data center or entirely in different data centers. It also enables entire virtual machines to be migrated from one host device to another. For an in-depth understanding of SDN-related technologies and landing practices, you can follow the official account of Kyushu Cloud Wechat and get the white paper on SDN plan. Many Kyushu Cloud SDN practices are analyzed in detail.

About Kyushu Cloud 99Cloud

Founded in 2012, Kyushu Cloud is the first professional company in China to engage in OpenStack and related open source services. Since its establishment six years ago, adhering to the concept of "open source enabling change", the company has grown from a single OpenStack product provider to an open source software and service provider covering cloud core, cloud operation, cloud transportation, cloud security and other fields. Kyushu Cloud has supported heavyweight customers such as State Grid, × ×, China UnionPay, China Mobile, China Telecom, China Unicom, China Resources Satellite, China Aviation Credit (Air Travel), eBay, International dry Port Group, China Life Insurance, Wanda Information, Dongfeng Motor, Nokia and other heavyweight customers. In the latest Queen release ranking in 2018, Kyushu Cloud ranks fourth in the world in core module contribution and second in China, with the world's first contribution in heavyweight projects such as container deployment Kolla project and NFV orchestration Tacker project.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.