CN108023814A - SDN control plane failure emergency systems and method - Google Patents

Abstract

The invention discloses an SDN control plane fault emergency system and method, wherein the system includes: single control plane and multiple controllers, used to control the emergency system to enter the single control plane and multiple controllers mode when a single point of failure occurs in the controller; The dual-plane hybrid in-band transmitter is used to control the emergency system to enter the dual-plane hybrid in-band transmission mode when there is a direct connection failure between some switches and the controller; the single data plane self-learning device is used to Control the emergency system to enter the single data plane self-learning mode, and extend the distributed routing protocol in the switch in a form compatible with the OpenFlow protocol, and expand the analysis of the protocol data packet to the action supported by the OpenFlow protocol, so as to cooperate with the flow entry completely with OpenFlow channels process packets for forwarding. The system can improve the controllability of the overall network and improve the stability and reliability of the SDN network.

Description

SDN control plane failure emergency system and method

technical field

The present invention relates to the technical field of network architecture, in particular to an SDN (Software Defined Networking, software defined network) control plane failure emergency system and method.

Background technique

As a new type of network architecture, SDN (Software Defined Networking) effectively solves the disadvantages of limited expansion of traditional network functions, and greatly improves the programmability and control capabilities of the network.

However, the centralized control architecture of SDN makes the status of the controller very special in the network. Once the data plane switch loses connection with the controller, it will affect the data transmission of the underlying network, and most SDN networks are deployed in the out-of-band mode of a single controller. , the deployment diagram is shown in Figure 1. The entire network is managed and controlled by only one controller (all switches are managed and controlled by the same controller), and there are direct links between the switches and the controller. In this deployment mode, the loss of connection between the switch and the controller is mainly divided into two situations: a single point of failure of the controller or a connection failure between the switch and the controller (including long-term link congestion, link physical failure, and link port failure). Wait).

When a single point of failure of the controller occurs, the failure of the controller will cause all switches to lose connection with the controller, the switch cannot actively generate the flow entries required for the underlying forwarding, and all data transmission on the network will be blocked before the controller recovers , the entire network is completely paralyzed;

In the event of a connection failure between the switch and the controller, there may be a connection failure between some switches and the controller, and the data transmission under the control of the lost switch is blocked. A controller single point of failure occurs similarly.

No matter what kind of failure occurs, the network will be greatly fluctuated when the switch is disconnected, the underlying host may not be able to communicate, and the entire network may even be completely paralyzed. This is unacceptable for commercial networks and needs to be resolved urgently.

Contents of the invention

The present invention aims to solve one of the technical problems in the related art at least to a certain extent.

Therefore, an object of the present invention is to propose an SDN control plane failure emergency system, which can reduce the impact of SDN control plane failures on the underlying network, and improve the stability and reliability of the SDN network.

Another object of the present invention is to propose an SDN control plane failure emergency method.

In order to achieve the above purpose, an embodiment of the present invention proposes an SDN control plane fault emergency system, including: a single control plane with multiple controllers, used to control the emergency system to enter the single control plane multiple controllers when a single point of failure occurs in the controller. Controller mode, and deploy multiple controllers on the control plane through redundant backup to solve the single point of failure of the controllers in the control plane; dual-plane hybrid in-band transmitters are used for direct connection between some switches and controllers When the connection fails, control the emergency system to enter the dual-plane hybrid in-band transmission mode, and transmit the control flow through other switches on the data plane, and the transmission process of the control flow is controlled by the controller, so as to pass the control The controller issues a flow table to manage the path of the control flow, so that all traffic in the network is directly controlled by the controller; the single data plane self-learning device is used to control the emergency system to enter the single data plane when the control plane fails. Plane self-learning mode, and extend the distributed routing protocol in the switch in a form compatible with the OpenFlow protocol, and extend the analysis of the protocol data packet to the action supported by the OpenFlow protocol, so as to cooperate with the flow entry completely in the OpenFlow channels process packets for forwarding.

The SDN control plane failure emergency system of the embodiment of the present invention can solve the problem of single point failure of the controller through a single control plane and multiple controllers; solve the problem of direct connection failure between some switches and controllers through a dual-plane hybrid in-band transmitter, Improve the controllability of the overall network; solve the problem of complete failure of the control plane through the single data plane self-learning device, effectively guarantee the normal communication of the underlying data plane, reduce the network recovery delay, improve network reliability, and reduce network management personnel's workload burden.

In addition, the SDN control plane failure emergency system according to the above-mentioned embodiments of the present invention may also have the following additional technical features:

Further, in one embodiment of the present invention, when a direct connection failure occurs between the part of the switches and the controller, the switch detects through the regularly sent ECHO data packets, and when no If the ECHO replies with a data packet, it is determined that the switch is disconnected from the controller, and the switch will try to re-establish a TCP (Transfer Control Protocol, Transmission Control Protocol) connection with the controller.

Further, in an embodiment of the present invention, the single data plane autonomous learner further includes: a processing module, the processing module is used to process the APR data packet by adding the entry, and by Adding an action realizes processing the OSPF data packet, wherein the action includes a table item of a special action; a self-learning module, the self-learning module is used for performing self-learning of the table item on the received data packet.

Further, in one embodiment of the present invention, the self-learning module further includes: a layer-2 entry self-learning unit, used to complete the learning of MAC (Medium Access Control, physical address) addresses and port numbers, to Instruct the final forwarding port of the data packet; the ARP entry self-learning unit is used to learn the corresponding relationship between the IP (Internet Protocol, Internet Protocol) address and the MAC address of the ARP data packet, so as to guide the host in other network segments to request the local network Forwarding of IP data packets between hosts under the segment; the self-learning unit of the three-layer table entry is used to calculate the next-hop switch to each target network segment in the network, so as to guide the IP between hosts under different network segments in the later stage packet forwarding.

Further, in one embodiment of the present invention, the self-learning module also includes:

The basic disaster recovery table entry unit is used to guide the data packet to realize jumping between different network layers.

In order to achieve the above object, another embodiment of the present invention proposes an SDN control plane fault emergency method, which includes the following steps: when a controller has a single point of failure, the control emergency system enters a single control plane multi-controller mode, and through Redundant backup deploys multiple controllers on the control plane to solve the single point of failure of the controllers on the control plane; when a direct connection failure occurs between some switches and controllers, the emergency system is controlled to enter the dual-plane hybrid zone transmission mode, and the control flow is transmitted through other switches in the data plane, and the transmission process of the control flow is controlled by the controller, so that the path of the control flow is managed by the controller through the flow table, so that all in the network The traffic is directly controlled by the controller; when the control plane fails, the emergency system is controlled to enter a single data plane self-learning mode, and the distributed routing protocol is extended in the switch in a form compatible with the OpenFlow protocol, and The parsing of protocol data packets is extended to actions supported by the OpenFlow protocol, so as to cooperate with the flow entries to completely process data packets through OpenFlow channels for forwarding.

The SDN control plane failure emergency method in the embodiment of the present invention can solve the problem of single point failure of the controller through a single control plane and multiple controllers; solve the problem of direct connection failure between some switches and controllers through a dual-plane hybrid in-band transmitter, Improve the controllability of the overall network; solve the problem of complete failure of the control plane through the single data plane self-learning device, effectively guarantee the normal communication of the underlying data plane, reduce the network recovery delay, improve network reliability, and reduce network management personnel's workload burden.

In addition, the SDN control plane failure emergency method according to the above-mentioned embodiments of the present invention may also have the following additional technical features:

Further, in an embodiment of the present invention, when a direct connection failure occurs between the part of the switches and the controller, it further includes: the switch detects through the ECHO data packet sent regularly, and when the specified time If the ECHO reply data packet is not received, it is determined that the switch is disconnected from the controller, and the switch will try to re-establish a TCP connection with the controller.

Further, in an embodiment of the present invention, after the control emergency system enters the single data plane autonomous learning mode, it further includes: processing the APR data packet by adding the entry, and implementing Processing the OSPF data packet, wherein the action includes an entry of a special action; performing self-learning of the entry on the received data packet.

Further, in an embodiment of the present invention, the self-study of the table entry for the received data packet further includes: completing the learning of the MAC address and port number, and guiding the final forwarding of the data packet port; learn the corresponding relationship between the IP address and the MAC address of the ARP data packet, and guide the hosts in other network segments to request the forwarding of IP data packets between the hosts in the local network segment; Next-hop switch, and guide the forwarding of IP data packets between hosts on different network segments in the later stage.

Further, in an embodiment of the present invention, the self-learning of the table entry for the received data packet further includes: guiding the data packet to realize jumping between different network layers.

Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.

Description of drawings

The above and/or additional aspects and advantages of the present invention will become apparent and easy to understand from the following description of the embodiments in conjunction with the accompanying drawings, wherein:

FIG. 1 is a schematic diagram of deployment of an SDN control plane failure emergency system in the related art;

FIG. 2 is a schematic structural diagram of an SDN control plane failure emergency system according to an embodiment of the present invention;

FIG. 3 is a flowchart of a layer 2 entry self-learning unit according to an embodiment of the present invention;

FIG. 4 is a flowchart of an ARP entry self-learning unit according to an embodiment of the present invention;

FIG. 5 is a flowchart of a three-layer entry self-learning unit according to an embodiment of the present invention;

Fig. 6 is a flow chart of distinguishing different data packets according to one embodiment of the present invention;

FIG. 7 is a schematic diagram of network deployment according to an embodiment of the present invention;

FIG. 8 is a schematic diagram of switching according to an embodiment of the present invention;

FIG. 9 is a flow chart of a method for emergency response to an SDN control plane failure according to an embodiment of the present invention.

Detailed ways

Embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the figures are exemplary and are intended to explain the present invention and should not be construed as limiting the present invention.

Before introducing the SDN control plane failure emergency system and method according to the embodiment of the present invention, the solution to the failure of the SDN controller in the related art will be introduced first.

For the situation of SDN controller failure, related technologies have two solutions, one is to start from the control plane, and the other is to start from the data plane.

(1) Multi-controller (standby controller)

Related technologies For the control plane, a solution of standby controller + data sharing center is proposed. The SDN network system includes a data sharing center and multiple controllers. Under normal working conditions, the main controller handles OpenFlow messages of network devices and manages network data flow, and the standby controller detects the status of the main controller and synchronizes information with it; controller status synchronization mainly includes: network view synchronization, controller network services Application state synchronization and controller data synchronization. For network view redundancy and synchronization, using the Hot active/standby strategy, the master controller will push real-time changes in the domain's network view to all standby controllers. For controller network service application redundancy and synchronization, using the Warm active/standby strategy, the primary controller periodically or conditionally pushes a snapshot of the service application to all standby controllers. For controller data synchronization, a reliable data sharing center is used between controllers.

The entire system is divided into two stages for the processing of faults: fault detection stage and recovery stage. Fault detection phase: The primary controller fault detection mechanism is based on the heartbeat component, and the main parameters include the interval of heartbeat messages and the time interval for the standby controller to repair the fault of the primary controller.

Recovery phase: When a failure of the main controller is detected, the system enters the recovery phase, which mainly includes the following steps:

First, designate a new primary controller, which is a standby controller with the highest ID (or IP);

Second, the new master controller notifies other controllers of their state changes;

Again, the recovery of controller network services and applications;

Finally, start the control network interface.

The overall system is based on the HAC (High Availability Controller) architecture. The main controller and the backup controller regularly send heartbeat messages to detect whether the controller fails. At the same time, the main controller and the backup controller will also follow the Different information types select different strategies for information synchronization. When the standby controller finds that the main controller fails, it enters the recovery phase and becomes the new main controller, and downloads part of the data information from the data sharing center to complete the synchronization of all data. Therefore, the new main controller takes over the entire network, solves the huge problem caused by the failure of a single controller, and can achieve lower synchronization overhead and 40ms to 50ms failure time, which can be accepted by network service applications.

(2) Mixed mode of data plane

In the SDN network environment, the flow entry of the switch is used to guide the forwarding of data packets. Generally, this part of the entry is delivered by the controller. Assuming that all controllers in the network fail or all direct connections between the controller and the switch fail, the switch cannot automatically generate the table entries required for forwarding. After the table entries issued by the controller fail, the underlying network will be completely paralyzed. Aiming at the extreme situation where the underlying forwarding device is completely out of the control of the controller, a solution based on traditional routing is proposed, and processing logic similar to traditional networks is added to the switch to ensure normal communication of the network.

For example, the dual-channel processing supported by OVS (Open vSwitch, software switch), that is, the OpenFlow processing channel and the traditional processing channel with some functions. The dual-channel mechanism allows the data packets entering OVS to enter the OpenFlow processing channel first, and then introduce the data packets into the traditional channel for processing through the special Normal port, and complete the functions of layer-2 forwarding and layer-3 routing according to the traditional protocol, so as to ensure the system in Forwarding and routing of data packets without a controller.

The OVS that is completely out of the control of the controller has two different working modes: failure independent mode and fail safe mode, and both of these two working modes can realize the forwarding function of data packets after the controller fails. Port flow entry to ensure the Layer 2 forwarding of data packets without a controller. The Normal channel implements Layer 2 forwarding according to the mapping relationship between MAC, PORT, and VLAN (Virtual Local Area Network, Virtual Local Area Network). Carry out simple learning while receiving the data packet, and store the source MAC address, ingress port and VLAN mapping relationship of the data packet. For the data packets received later, if the destination MAC address is found in the storage table of the Normal channel, it will be directly forwarded from the corresponding port, and if the search fails, it will be broadcast and forwarded.

In the failure-independent mode, after the switch actively detects the connection status of the controller three times and finds that the connection fails, ovs-vswitchd will automatically take over the forwarding logic of the device (still trying to connect to the controller in the background, and exit this mode once the connection is restored). In this mode, after it is determined that the connection is lost, it will automatically issue a hidden entry with the highest priority, any matching domain, and Normal port. At this time, OVS will act as an ordinary MAC-learning switch and will receive the received data. The packet is switched to the Normal channel to ensure Layer 2 forwarding.

In the fail-safe mode, the switch will discard the data packets or messages uploaded to the controller (the background still tries to connect to the controller, and exits this mode once the connection is restored). In this mode, entries containing Normal ports will not be automatically configured, and OVS will only forward according to the existing entries, and the original entries in the flow table will be deleted according to the timeout of hard_timeout and idle_timeout. Therefore, in this mode, if you want to ensure Layer 2 communication between hosts, you need to download the entry in advance. The hard_timeout and idle_timeout of the entry are both 0, the port is Normal, and both times are 0 to ensure that the entry will not time out. The matched data packets are introduced into the Normal channel to ensure simple Layer 2 forwarding. The disadvantage of "fail-safe mode" compared with "failure-independent mode" is that entries containing Normal need to be delivered in advance.

However, the applicable scenarios of related technologies are limited, that is to say, related technologies are only applicable to solve the single point of failure of the controller or the situation where all switches are separated from the controller, and the situation of direct connection failure of some switches is not considered, and the solution cannot be smoothly transitioned , the lack of coping strategies for different fault levels, and limited application scenarios; followed by the lack of an overall system, the related technology only starts from a certain plane alone, the multi-controller mode solves the problem from the control plane, and the hybrid mode solves the problem from the data plane , the two methods are independent of each other. However, the faults that may occur in a real network environment are often complex and changeable, and may be caused by a mixture of various situations. There is not yet a mature system, and a single solution cannot fully adapt to the changing and complex real network environment; thirdly, the compatibility is poor. The OpenFlow protocol needs to be developed, and the OpenFlow-hybrid switch needs to be supported. At present, there are few hybrid switches that support both OpenFlow channels and traditional processing channels with normal ports, and the complexity is high, and the price of the equipment is also high. In addition, adding Normal channel processing in the SDN network is contrary to the SDN idea of separation of control and forwarding, and does not have scalability and flexibility. Finally, communication between different network segments is impossible. In the promiscuous mode solution, even if OVS that supports the normal function, when disconnecting the controller, the normal channel of OVS itself does not support any routing protocol, and cannot self-learn the network topology after disconnecting the controller. This will only support communication between the same network segment after the controller is disconnected, that is, simple Layer 2 forwarding, which limits the expansion of the network scale and reduces the application scenarios.

Based on the above problems, the present invention proposes an SDN control plane failure emergency system and method.

The following describes the SDN control plane failure emergency system and method according to the embodiments of the present invention with reference to the drawings. First, the SDN control plane failure emergency system according to the embodiments of the present invention will be described with reference to the drawings.

Fig. 2 is a schematic structural diagram of an SDN control plane failure emergency system according to an embodiment of the present invention.

As shown in FIG. 2 , the SDN control plane fault emergency system 10 includes: a single control plane multi-controller 100 , a dual plane hybrid in-band transmitter 200 and a single data plane autonomous learner 300 .

Among them, the single control plane multi-controller 100 is used to control the emergency system to enter the single control plane multi-controller mode when a controller has a single point of failure, and deploy multiple controllers on the control plane through redundant backup to solve the problem of control plane failure. controller single point of failure. The dual-plane hybrid in-band transmitter 200 is used to control the emergency system to enter the dual-plane hybrid in-band transmission mode when there is a direct connection failure between some switches and the controller, and transmit the control flow through other switches on the data plane, and the transmission of the control flow The process is controlled by the controller, and the path of the control flow is managed through the controller issuing the flow table, so that all traffic in the network is directly controlled by the controller. The single data plane autonomous learner 300 is used to control the emergency system to enter the single data plane autonomous learning mode when the control plane fails, and extend the distributed routing protocol in the switch in a form compatible with the OpenFlow protocol, and analyze the protocol data packets It is extended to the action supported by the OpenFlow protocol, so as to cooperate with the flow entry to completely process the data packet through the OpenFlow channel for forwarding. The system 10 of the embodiment of the present invention can reduce the impact of SDN control plane failure on the underlying network, and improve the stability and reliability of the SDN network.

It can be understood that the system of the embodiment of the present invention may include a single control plane multi-controller 100, a dual-plane hybrid in-band transmitter 200, and a single data plane autonomous learner 300, so as to be able to smooth between stages according to different fault conditions switch.

Wherein, the single control plane multi-controller 100 deploys multiple controllers on the control plane through redundant backup, so as to solve the single point of failure of the controller of the control plane. For example, by selecting one of the controllers as the main controller to manage and control the network, and synchronize information between the controllers. When the main controller fails, a new main controller will be elected from the standby controllers, and all services will be switched to the new main controller. The switching time is very short and completely imperceptible on the data plane. The single control plane multi-controller 100 in the embodiment of the present invention may adopt a technical solution in the related art, and in order to reduce redundancy, details are not described here.

The dual-plane hybrid in-band transmitter 200 considers not only the single-point failure of the controller, but also the physical failure of some control links or the failure of link ports on the control plane, resulting in the loss of connection between some switches and the controller. That is to say, when the controller is connected normally, the controller issues a flow entry with a specific priority and a specific matching item according to the characteristics of the control flow, so as to guide the control flow to be forwarded from the specified direct connection port to the controller. The switch actively detects the connection status with the controller. If the connection with the controller is interrupted, it will actively delete the flow entry in the current switch that guides the forwarding of the control packet, and forward the control packet according to the disaster recovery table issued in advance. Initially, the switch floods the control packets from all ports to the surrounding switches and forwards them to the controller via the surrounding switches. If multiple paths can reach the controller, the fastest reply path will be selected as the fixed path forwarding (but the fixed path may not be the shortest path). After the switch re-establishes the connection with the controller through data plane forwarding, the controller can calculate the optimal path between the switch and the controller at this time, and issue flow entries to guide subsequent control packets to be uploaded to the controller via the optimal path ( The optimal path here is determined by the developer, and applications are developed on the controller to implement related algorithms, which are not specifically limited here).

The single data plane autonomous learner 300 mainly considers extreme situations, such as a situation where all controllers fail or all control links fail, resulting in the loss of connection between all switches and controllers. That is to say, after losing connection with the controller, the switch can automatically generate flow entries locally to replace the function of the controller. The working mode of the switch at this time is similar to that of the switch in the network in the related art, but it does not use the data channel of the related technology to process the data packet, but processes the data packet in the form of OpenFlow flow table matching. In the embodiment of the present invention, the switch actively detects the connection status with the controller. If the connection with the controller is interrupted, the disaster recovery table item is sent actively, and the distributed routing protocol RIP protocol or OSPF protocol starts to be executed. Unlike hybrid switches that use channels in related technologies to process data packets, all data packets (including RIP protocol data packets or OSPF protocol data packets) are processed by OpenFlow channels, and the analysis of protocol data packets is extended to actions supported by OpenFlow , by matching protocol data packets and performing corresponding parsing actions, it will be able to autonomously generate flow entries to guide the forwarding and routing of data packets between different network segments.

Further, in one embodiment of the present invention, when a direct connection failure occurs between some switches and the controller, the switch detects through the ECHO data packets sent regularly, and when the ECHO reply data packet is not received within the specified time, then It is judged that the switch is disconnected from the controller, and the switch will try to re-establish a TCP connection with the controller.

It is understandable that in an SDN network system, when the direct connection between some switches and the controller fails, the control packets sent by the switch cannot communicate with the controller through the original direct connection port. The switch will be disconnected from the controller. In order to ensure that the switch can detect the connection status between itself and the controller in real time, the embodiment of the present invention adopts a corresponding strategy for fault recovery. That is to say, the switch in the embodiment of the present invention can send The ECHO data packet is used for detection. When the ECHO reply data packet is not received within the specified time (this time will be dynamically changed according to the actual network link delay), it will judge that it is disconnected from the controller, and the switch will try to reconnect with the controller. Establish a TCP connection.

The failure of the direct connection makes the controller unreachable through the direct link, so after the switch detects the disconnection, the control flow will match the disaster recovery uplink entry (as shown in Table 1), and the flooding method is used initially Establish a connection with the controller, and the sent control packets will reach all adjacent switches, and the adjacent switches contain uplink entries to the controller (as shown in Table 2), and the matching domain is the IP address of the controller and the TCP destination port No., the action is to forward from a certain port (assuming that the neighbor switch is reachable to the controller, it may reach the controller through a direct link, or through its own neighbor switch). Through forwarding by adjacent switches, the controller can receive the control packet from the switch with link failure. Wherein, Table 1 is the uplink entry sent by the switch, and Table 2 is the uplink entry sent by the controller.

Table 1

Table 2

After the controller receives the control packet from the switch, it forwards the reply data packet to the requesting switch along the reverse path of the uplink path. Although the control packet sent by the switch for the first time is broadcast to the surrounding switches, once a path to the controller is established, this path will be temporarily used. In the process of re-establishing the TCP connection, when the switch receives the control packets from the surrounding switches, it will generate OpenFlow entries through port learning to record the mapping relationship between MAC addresses and ports (as shown in Table 3), which is equivalent to recording control The upstream path of the packet, so as to ensure that the downstream traffic of the controller can send the data packet to the requesting switch along the correct unicast path (as shown in Table 4 and Table 5). In addition, after the requesting switch receives the control packet replied by the controller, it will generate an OpenFlow entry to the controller. All upstream traffic arriving at the controller will match this entry, preventing the switch from communicating with the controller through broadcasting. In this way, the switch whose direct connection fails can re-establish a connection with the controller and accept the control of the controller. Among them, Table 3 is the port learning entry recorded by the switch autonomous learning, Table 4 is the downlink entry sent by the switch, and Table 5 is the downlink entry sent by the controller.

table 3

Destination MAC address action (The source MAC address of the packet received by the switch) Outgoing port = (the inbound port where the switch receives the data packet)

Table 4

table 5

After the switch normally establishes a connection with the controller through the disaster recovery uplink and downlink entries issued by itself, the controller will calculate the optimal path between the controller and the requesting switch control packet according to the real-time connection status in the network, and send the request to the switch and The switch that needs the route issues the relevant table items, so as to ensure that in this mode, the controller can still manage the route selection of the control packet and reduce the delay of the control packet. In the embodiment of the present invention, in this way, the switch whose direct connection fails can re-establish a connection with the controller and accept the management and control of the controller.

The controller always starts the control flow management application, and issues table entries for managing control flow to each switch normally connected to the controller. The controller application will judge the optimal path from the current switch to itself in real time according to the obtained switch port information and current network topology information. When the direct connection path is normal, the next entry will be forwarded through the direct connection path. , the controller can select a path for the switch according to a certain strategy (the strategy is determined by the upper-layer application developer of the controller, and no specific limitation is made here).

The control flow management application in the embodiment of the present invention can be triggered by a specific Packet_in data packet, and the switch sends a specific Packet_in data packet to the controller at regular intervals, and the content of the data packet includes the MAC address, DPID and identifier of the switch. The controller can identify the Packet_in data packet through the identifier, trigger the control flow management application, read the current network information and switch port information, select the optimal path for the request switch control flow, and publish it to all involved switches on the path Items, wherein, the structure of the uplink entry is consistent with Table 2, and the structure of the downlink entry is consistent with Table 4.

In addition, during the process from the moment the switch is disconnected to re-establishing a connection with the controller, the original data plane data packet forwarding flow entry in the switch will still exist, ensuring that the original data forwarding of the data plane will not be affected during this process. However, after the connection is restored, the controller re-delivers the data plane entry to minimize the impact of the connection failure on the underlying communication.

In the dual-plane hybrid in-band transmitter 200, the switch whose direct connection fails can still perform OpenFlow communication with the controller through the adjacent switch, and the controller will dynamically adjust the path between itself and the switch according to the real-time changes in the network , to ensure low latency of control traffic, improve service levels, and at the moment of disconnection, the switch retains the original data plane entries, maintains the original communication, and reduces the impact of faults on data packet forwarding.

The dual-plane hybrid in-band transmitter 200 in the embodiment of the present invention can solve the problem that the switch loses the control of the controller due to the failure of some control connections, and enables the entire system to perform smooth fault recovery under different degrees of faults on the control plane, ensuring that any degree of fault , can restore the normal communication of the network in a relatively short period of time, and improve the reliability of the network.

Further, in an embodiment of the present invention, the single data plane self-learning device 300 further includes: a processing module and a self-learning module. Wherein, the processing module is used to process the APR data packet by adding an entry, and realize processing the OSPF data packet by adding an action, wherein the action includes a special action entry. The self-learning module is used for self-learning of table items for received data packets.

It can be understood that the OpenFlow protocol is a commonly used southbound protocol in the SDN architecture. Based on the multi-level flow table mechanism proposed by the OpenFlow protocol, a selective processing scheme for the multi-level flow table is designed. In one embodiment of the present invention, once the data packet enters the switch, the basic disaster recovery entry guides the data packet into the type discrimination table to determine the type of the data packet, and guides it to the corresponding processing table according to the target type. In the processing flow of each type of data packet, the design of a multi-level flow table may also be used to achieve the purpose of correct handling of different situations.

Since the underlying network has been completely separated from the control plane, similar to the network in the related art, the routes of the entire network must be obtained between switches through a distributed routing protocol. In the network of the related technology, before the IP data packet communication between the hosts, the ARP data packet and the OSPF data packet must have appeared in the network (or other routing protocols, such as RIP, OSPF, ISIS, etc., this part takes OSPF as an example) In the embodiment of the present invention, by analyzing the basic requirements of network data packet forwarding, it is found that only ARP and OSPF data packets need to be learned and correctly processed to achieve the expected goal.

In the embodiment of the present invention, the single data plane self-learning device 300 can be divided into a processing module and a self-learning module. The processing module can correctly process the received data packet: for example, directly process the ARP data packet by adding an entry; and realize the processing of the OSPF data packet by adding an action. The self-learning module can perform table entry self-learning on the received data packets. And the self-learning module is the key point of the single data plane self-learning device 300, which mainly divides the modules according to the network level. The unit and the three-layer entry self-learning unit are not described in detail here in order to reduce redundancy.

Furthermore, in an embodiment of the present invention, the self-learning module further includes: a layer-2 table entry self-learning unit, an ARP table entry self-learning unit, and a layer-3 table entry self-learning unit. Among them, the two-layer table item self-learning unit is used to complete the study of the MAC address and port number to guide the final forwarding port of the data packet; the ARP table item self-learning unit is used to learn the IP address and MAC address of the ARP data packet The corresponding relationship of addresses is used to guide hosts in other network segments to request the forwarding of IP data packets between hosts in the local network segment. The three-layer entry self-learning unit is used to calculate the next-hop switch to each target network segment in the network, so as to guide the forwarding of IP data packets between hosts in different network segments in the later stage.

It can be understood that the layer 2 entry self-learning unit mainly completes the learning of the MAC address and the port number, and is used to guide the final forwarding port of the data packet. The basic principle of the design of the self-learning unit of the two-layer table entry is very close to the working principle of the bridge in the related technology network, and records the source MAC address and the ingress port of the data packet when it enters the bridge. , at the same time may include a three-layer network, combined with the actual situation of the port, so only the port is learned for ARP and OSPF packets. Specifically, as shown in Figure 3, the switch first analyzes the key field in the data packet for the received ARP and OSPF data packets, extracts the source MAC address of the data packet and records the port number entering the switch, and then according to The information constructs a new OpenFlow table entry. The matching field of the table entry is the destination MAC address. The value of the matching field is the source MAC address of the data packet. The action performed is forwarding from the specified port, and the forwarding port is the incoming port of the data packet. Finally, it is judged whether the entry already exists in the specified flow table, if it exists, the original entry is updated, and if it does not exist, the entry is added to the specified flow table.

The ARP table entry self-learning unit mainly learns the correspondence between the IP address and the MAC address of the ARP data packet. The table entries learned by the ARP table entry self-learning unit are mainly used to guide hosts on other network segments to request hosts on the local network segment. Forwarding of IP data packets between them. The idea of self-learning is basically similar to the learning process of the second layer, except that the content of learning becomes the corresponding relationship between IP addresses and MAC addresses. The processing flow of ARP data packets is shown in Figure 4. For the received ARP request packets and reply packets, the switch first parses the packet headers, extracts the source IP address and MAC address of the data packets, and then constructs a new The OpenFlow table item, the table item matching field is the use of IP protocol and destination IP, the value of the matching field is the source IP address of the data packet, the action to be executed is to modify the destination MAC of the data packet to the source MAC address of the ARP data packet, and finally Judging whether the entry already exists in the specified flow table, if it exists, update the original entry, and if it does not exist, add the entry to the specified flow table.

The main goal of the three-layer entry self-learning unit is to calculate the next-hop switch to each target network segment in the network, and issue corresponding entries. This part of the entries is used to later guide the forwarding of IP data packets between hosts on different network segments. As shown in Figure 5, the design idea of the three-layer entry self-learning unit is mainly to extract the content of the OSPF link state update message (LSU) and store it in the database, and encapsulate the three-layer forwarding entry according to the synchronized database content . Therefore, this process requires switches to support receiving and sending OSPF data packets. OVS analyzes the data packets and extracts link state information and stores them in the local database. In each switch, the Dijkstra algorithm is performed according to the content of the database to calculate the The next-hop switch of the shortest path for each target network segment, and read the target network segment, the next-hop switch port MAC address leading to the target network segment, and the corresponding forwarding port MAC address of the local switch according to the database content, and then based on these information Construct a new OpenFlow entry. The matching field of the entry is the use of the IP protocol and the destination network segment. The value of the matching field is an address with a subnet mask length. The action to be executed is to modify the destination MAC of the data packet to the next hop Switch port MAC address, modify the source MAC address of the data packet to the forwarding port MAC address. Finally, it is judged whether the entry already exists in the specified flow table, if it exists, the original entry is updated, and if it does not exist, the entry is added to the specified flow table.

Further, in an embodiment of the present invention, the self-learning module further includes: a basic disaster recovery entry unit. Among them, the basic disaster recovery entry unit guides data packets to realize jumping between different network layers.

It is understandable that, as shown in Figure 6, all data packets in OVS will first enter table 0 for matching search by default. Add an entry here to judge the connection status of the controller. If the controller is normal, directly Use the table items issued by the controller for forwarding. If the controller fails, it will jump to the type judgment table. It should be noted that the priority of the connection status judging entry should be higher than that uploaded to the controller, so that the connection status can be judged first. The table entered after the jump in the previous step mainly judges the data packet type, distinguishes ARP data packets, IP data packets communicated between hosts and OSPF data packets communicated between switches, and adds corresponding jumps to these different types of data packets. Forward, forward, discard or self-learning actions. Among them, the processing of ARP data packets is divided into two parts: firstly, the switch acts as a gateway when requesting the MAC address of the gateway and directly replies to the ARP request according to the entry; segment of the ARP reply packet, it will directly look up the entry for unicast forwarding. In addition, each table will issue a completely matching entry, and the execution action will be set to discard, broadcast, or other according to the specific situation. Finally, you need to add the table Item to reduce the TTL value of IP packets between hosts.

In addition, based on the connection state detection mechanism in OVS, a scheme for publishing entries based on the connection state is designed. When the state value changes, the disaster recovery entry action will be triggered, and all disaster recovery entries will be automatically deleted when the connection with the controller is restored. Standby table entries, that is, only send this part of table entries when the controller is disconnected, and this part of table entries will not appear when the controller is normal, thereby reducing the processing burden in the normal connection state.

In the basic disaster recovery entry unit, even if all switches are out of the control of the controller, due to the distributed routing protocol running on them and some basic table entry learning, it is guaranteed that the underlying devices can still communicate normally in this state , to reduce the impact of faults on packet forwarding.

For example, in a specific embodiment of the present invention, the application scenario of the SDN control plane failure emergency system of the embodiment of the present invention is as follows:

When different failures occur on the control plane, the system can smoothly switch between levels. The whole process does not require manual intervention and is completely completed by the system to ensure normal transmission of the data plane and greatly reduce the impact of control plane failures on the data plane.

As shown in FIG. 7 , the embodiment of the present invention deploys multiple controllers on the control plane, and each controller has a direct link with each switch, so that normal OpenFlow communication can be established. (The meaning of this direct connection is that there is a path on the network outside the data plane. In fact, they are not necessarily directly connected. They may be forwarded and connected by many existing network switches, or they may be connected by only one switch. According to Actual deployments vary, so clouds are used in illustrations to cover all cases.)

The embodiment of the present invention does not discuss possible link failures in the cloud, but only discusses link failures of switches accessing the cloud.

In the initial mode, the network is working normally, the control plane will elect a certain controller as the master controller, and the other controllers are the backup controllers. Assume that C1 in the schematic diagram is the current master controller, as shown in the sub-diagram of Figure 7 As shown in A, in the normal working mode, the main controller C1 controls the transmission of the data plane.

Assume that the current master controller C1 fails. As shown in sub-diagram B of FIG. 7 , the backup controllers C2 and C3 will detect the failure of the master controller and elect a new master controller, assumed to be C2. C2 will download and synchronize the content of C1 from the data sharing center, complete the complete switchover of the controller work, and inform the switch that it has become the new master controller, and successfully take over the data transmission of the bottom switch (as shown in ① in Figure 8 ), where the controller scheme and deployment mode are not fixed, and can be selected according to actual needs during network deployment, and no specific restrictions are set here.

After the new main controller C2 takes over the network, it will start the control flow management application (when C1 is working normally, it will also start the application). The application can judge the failure status of the current switch port and the link to which the port belongs in real time according to the port information uploaded by the switch. If it is in normal working condition, the application will issue a flow table to guide the control flow to be forwarded from the directly connected port. The control flow transmission at this time belongs to out-of-band transmission, that is, each switch transmits the control flow through a dedicated non-data plane network.

In the multi-controller mode of single control plane, the delay of switching to the standby controller after a single point failure of the controller is extremely short, and it is completely unaware on the data plane. Therefore, when the controller has a single point of failure, the data transmission of the data plane is not affected at all. .

It is assumed that the control link of the current switch S1 has a physical failure (a link port failure or long-term congestion). As shown in subgraph C of FIG. 7 , its control flow will not be normally forwarded from the link to the controller, so the switch S1 will detect that the connection with the controller is interrupted.

The switch will actively delete the flow entry sent by the controller application to direct the control flow to be forwarded from the directly connected port (the existence of the flow entry will always guide the control flow to be forwarded to the faulty link, so it must be deleted), and according to The pre-delivered disaster recovery flow entry forwards the control packet. At this time, because it is not controlled by the controller, the switch will randomly select a path to the controller in the data plane to forward the control packet. Assume that the control flow of the current switch S1 is forwarded to the controller via the switch S2 and re-established with the controller. Connection, after the connection with the controller is re-established, the control flow management application will take over the forwarding of the control flow again, and select a new forwarding path for the control flow of the switch according to the current link failure and other factors. The control flow transmission at this time belongs to in-band transmission, that is, the switch transmits the control flow through the network of the data plane.

In the dual-plane hybrid in-band transmitter, during the switching from out-of-band to in-band, the switch will lose connection with the controller for a period of time. In order to continue to ensure the normal transmission of the underlying network, the After the connection, it will enter the single data plane self-learning mode, that is, the self-learning unit (as shown in ② in Figure 8), and start to automatically generate table entries to guide data forwarding. Since it takes a certain amount of time to converge to generate entries through switch self-learning, the existing entries in the switch issued by the controller will not be deleted and can continue to guide the forwarding of data packets. When the entries are deleted over time, the Execute the entries generated by the switch self-learning. Using the dual protection mechanism of reserved entries under the controller and self-learning entries greatly reduces the impact on the underlying data plane during the switching process. After the switching is completed, the controller will take over the switch again, and the switch will automatically exit the single data plane self-learning mode and enter the in-band management mode (as shown in ③ in the transition diagram).

Assume that all switches currently control link physical failure (link port failure or long-term congestion). As shown in subgraph D of FIG. 7 , no control flow of any switch has a fault-free path to the controller, so all switches will detect that they are disconnected from the controller.

At this time, the actions of the switch are exactly the same as when switching the in-band mode. The only difference is that when there is a normal working control link, it can successfully switch to the in-band mode. Single data plane self-learning mode (as shown in ④ in Figure 8).

Assuming that the control link of some switches is restored, because the background of the switch is still trying to establish a connection with the controller in the single data plane self-learning mode, it will re-establish a connection with the controller through the restored control link, and the controller will After taking over the switch again, the switch will automatically exit the single data plane self-learning mode and enter the in-band management mode (as shown in ⑤ in Figure 8). The process of re-establishing the connection is exactly the same as when switching the in-band mode, and will not be repeated here to reduce redundancy.

Assuming that the current control links of all switches are restored, the control flow management application on the controller will judge that all links are restored to normal based on the port information uploaded by the switch. Therefore, the control flow management application will update the guidance control flow in all switches. For the flow entries sent, select the direct connection path for all control flows to forward, and restore the out-of-band mode (as shown in ⑥ in Figure 8).

After the above process, the entire process from failure to recovery is completed. The whole process is completely completed by the switch and the controller independently. In the whole process, the normal progress of the underlying data transmission is always guaranteed to the greatest extent, which greatly reduces the impact of various failures. The impact of the underlying network. The above process only describes a fault occurrence sequence and recovery sequence, but the system is not limited to solving faults in this sequence, and the system can work normally in any fault sequence. In addition, all controller failures are the same as all control link failures. To reduce redundancy, details are not described here.

In summary, the SDN control plane failure response system in the embodiment of the present invention has the following advantages:

(1) Wide range of applications: The SDN control plane failure emergency system has designed different modes for control plane failures, which can adapt to various control plane failures or any combination of failures, and is applicable to the deployment of any SDN network.

(2) Improve network reliability: The SDN control plane failure emergency system can smoothly switch between levels, minimize the impact of control plane failures, and ideally achieve seamless network switching and improve network reliability.

(3) Guarantee the communication after some switches lose connection: After some switches and the controller are disconnected, they can switch to the in-band mode, and use the data plane switch to forward the control packet, so that the controller can re-establish a connection with the switch and maintain the underlying data Flat communication.

(4) Improve the controllability of the network: the controllable in-band management module is adopted to realize the real-time control of the in-band control flow by the controller, and the corresponding strategy can be applied to the forwarding of the control flow. Improve the controllability of the network, in line with the idea of SDN.

(5) Guarantee the communication after the complete failure of the control plane: the three-layer self-learning mode is adopted. After the complete failure of the control plane, the switch can independently generate three-layer routing entries to ensure the normal communication of the network.

(6) Expand the scope of network deployment: the switch is suitable for the three-layer network, which expands the scope of network deployment.

According to the SDN control plane failure response system proposed by the embodiment of the present invention, the problem of single point failure of the controller can be solved by using a single control plane and multiple controllers; the problem of direct connection failure between some switches and controllers can be solved by using a dual-plane hybrid in-band transmitter problem, improving the controllability of the overall network; solving the problem of complete failure of the control plane through a single data plane self-learning device; The normal communication of the plane reduces the delay of network recovery and improves the reliability of the network; the switching of the system does not require manual intervention, reducing the burden of network management personnel.

Next, the SDN control plane failure emergency method proposed according to the embodiment of the present invention will be described with reference to the accompanying drawings.

FIG. 9 is a flow chart of a method for emergency response to an SDN control plane failure according to an embodiment of the present invention.

As shown in Figure 9, the SDN control plane failure emergency method includes the following steps:

In step S901, when a controller has a single point of failure, the control emergency system enters a single control plane multi-controller mode, and deploys multiple controllers on the control plane through redundant backup, so as to solve the controller single point failure of the control plane.

In step S902, when there is a direct connection failure between some switches and the controller, the control emergency system enters the dual-plane hybrid in-band transmission mode, and transmits the control flow through other switches on the data plane, and the transmission process of the control flow is controlled by the controller. Control, to manage the path of the control flow through the controller issuing the flow table, so that all traffic in the network is directly controlled by the controller.

In step S903, when the control plane fails, the control emergency system enters the single data plane autonomous learning mode, and the distributed routing protocol is extended in the switch in a form compatible with the OpenFlow protocol, and the analysis of the protocol data packet is extended to the OpenFlow protocol Supported actions, in order to cooperate with the flow entry to completely process the data packet through the OpenFlow channel for forwarding.

Further, in one embodiment of the present invention, when a direct connection failure occurs between some switches and the controller, it further includes: the switch detects through the ECHO data packets sent regularly, and when the ECHO reply data is not received within the specified time packet, it is determined that the switch is disconnected from the controller, and the switch will try to re-establish a TCP connection with the controller.

Further, in one embodiment of the present invention, after controlling the emergency system to enter the single data plane autonomous learning mode, it further includes: processing the APR data packet by adding an entry, and processing the OSPF data packet by adding an action ; Perform entry self-learning on the received data packet, wherein the action includes a special action entry.

Further, in one embodiment of the present invention, performing table entry self-study on the received data packet further includes: completing the learning of MAC address and port number, and guiding the final forwarding port of the data packet; learning the ARP data packet The corresponding relationship between the IP address and the MAC address, and guide the hosts in other network segments to request the forwarding of IP data packets between the hosts in the local network segment; calculate the next-hop switch to each target network segment in the network, and In the later stage, it guides the forwarding of IP data packets between hosts on different network segments.

Further, in an embodiment of the present invention, performing table entry self-learning on the received data packets also includes: guiding the data packets to realize jumping between different network layers.

It should be noted that, the foregoing explanations of the embodiment of the SDN control plane failure emergency system are also applicable to the SDN control plane failure emergency method of this embodiment, which will not be repeated here.

According to the SDN control plane failure emergency method proposed in the embodiment of the present invention, the problem of single point failure of the controller can be solved by using a single control plane and multiple controllers; the problem of direct connection failure between some switches and controllers can be solved by using a dual-plane hybrid in-band transmitter problem, improving the controllability of the overall network; solving the problem of complete failure of the control plane through a single data plane self-learning device; Plane normal communication reduces network recovery delay, improves network reliability, and reduces the burden on network management personnel.

In describing the present invention, it should be understood that the terms "center", "longitudinal", "transverse", "length", "width", "thickness", "upper", "lower", "front", " Back", "Left", "Right", "Vertical", "Horizontal", "Top", "Bottom", "Inner", "Outer", "Clockwise", "Counterclockwise", "Axial", The orientation or positional relationship indicated by "radial", "circumferential", etc. is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying the referred device or element Must be in a particular orientation, be constructed in a particular orientation, and operate in a particular orientation, and therefore should not be construed as limiting the invention.

In addition, the terms "first" and "second" are used for descriptive purposes only, and cannot be interpreted as indicating or implying relative importance or implicitly specifying the quantity of indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include at least one of these features. In the description of the present invention, "plurality" means at least two, such as two, three, etc., unless otherwise specifically defined.

In the present invention, unless otherwise clearly specified and limited, terms such as "installation", "connection", "connection" and "fixation" should be understood in a broad sense, for example, it can be a fixed connection or a detachable connection , or integrated; it may be mechanically connected or electrically connected; it may be directly connected or indirectly connected through an intermediary, and it may be the internal communication of two components or the interaction relationship between two components, unless otherwise specified limit. Those of ordinary skill in the art can understand the specific meanings of the above terms in the present invention according to specific situations.

In the present invention, unless otherwise clearly specified and limited, the first feature may be in direct contact with the first feature or the first and second feature may be in direct contact with the second feature through an intermediary. touch. Moreover, "above", "above" and "above" the first feature on the second feature may mean that the first feature is directly above or obliquely above the second feature, or simply means that the first feature is higher in level than the second feature. "Below", "beneath" and "beneath" the first feature may mean that the first feature is directly below or obliquely below the second feature, or simply means that the first feature is less horizontally than the second feature.

In the description of this specification, descriptions referring to the terms "one embodiment", "some embodiments", "example", "specific examples", or "some examples" mean that specific features described in connection with the embodiment or example , structure, material or characteristic is included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms are not necessarily directed to the same embodiment or example. Furthermore, the described specific features, structures, materials or characteristics may be combined in any suitable manner in any one or more embodiments or examples. In addition, those skilled in the art can combine and combine different embodiments or examples and features of different embodiments or examples described in this specification without conflicting with each other.

Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present invention, those skilled in the art can make the above-mentioned The embodiments are subject to changes, modifications, substitutions and variations.

Claims (10)

1. A kind of SDN control plane failure emergency system, is characterized in that, comprises the following steps: Single control plane with multiple controllers is used to control the emergency system to enter the single control plane with multiple controllers mode when a controller has a single point of failure, and deploy multiple controllers on the control plane through redundant backup to solve the problem of the control plane. The controller is a single point of failure; The dual-plane hybrid in-band transmitter is used to control the emergency system to enter the dual-plane hybrid in-band transmission mode when a direct connection failure occurs between some switches and the controller, and transmit the control flow through other switches on the data plane, and the The transmission process of the control flow is controlled by the controller, so as to manage the path of the control flow through the controller issuing a flow table, so that all traffic in the network is directly controlled by the controller; and The single data plane self-learning device is used to control the emergency system to enter the single data plane self-learning mode when the control plane fails, and extend the distributed routing protocol in the switch in a form compatible with the OpenFlow protocol, and the protocol The parsing of the data packet is extended to an action supported by the OpenFlow protocol, so as to cooperate with the flow entry to completely process the data packet through the OpenFlow channel for forwarding. 2. The SDN control plane failure emergency system according to claim 1, wherein when a direct connection failure occurs between the part of the switches and the controller, the switch detects by the ECHO packets sent regularly , when the ECHO reply data packet is not received within a specified time, it is determined that the switch is disconnected from the controller, and the switch will try to re-establish a TCP connection with the controller. 3. SDN control plane fault emergency system according to claim 1, is characterized in that, described single data plane autonomous learner, further comprises: A processing module, the processing module is used to process the APR data packet by adding the table item, and realize processing the OSPF data packet by adding an action, wherein the action includes a special action table item; A self-learning module, the self-learning module is used to perform self-learning of the entry on the received data packet. 4. SDN control plane failure emergency system according to claim 3, is characterized in that, described self-learning module, further comprises: The self-learning unit of the two-layer table entry is used to complete the learning of the MAC address and the port number, so as to guide the final forwarding port of the data packet; The ARP entry self-learning unit is used to learn the corresponding relationship between the IP address and the MAC address of the ARP data packet, so as to guide the hosts in other network segments to request the forwarding of IP data packets between the hosts in the local network segment in the later stage; The three-layer entry self-learning unit is used to calculate the next-hop switch to each target network segment in the network, so as to guide the forwarding of IP data packets between hosts in different network segments in the later stage. 5. SDN control plane fault emergency system according to claim 1, is characterized in that, described self-learning module, also comprises: The basic disaster recovery table entry unit is used to guide the data packet to realize jumping between different network layers. 6. A kind of SDN control plane failure emergency method, is characterized in that, comprises the following steps: When a controller has a single point of failure, the control emergency system enters a single control plane multi-controller mode, and deploys multiple controllers on the control plane through redundant backup to solve the controller single point failure of the control plane; When there is a direct connection failure between some switches and the controller, control the emergency system to enter the dual-plane hybrid in-band transmission mode, and transmit the control flow through other switches on the data plane, and the transmission process of the control flow is controlled by the controller control, so that the path of the control flow is managed by issuing a flow table through the controller, so that all traffic in the network is directly controlled by the controller; and When the control plane fails, the emergency system is controlled to enter the single data plane autonomous learning mode, and the distributed routing protocol is extended in the switch in a form compatible with the OpenFlow protocol, and the analysis of the protocol data packet is extended to the OpenFlow The action supported by the protocol is to cooperate with the flow entry to completely process the data packet through the OpenFlow channel for forwarding. 7. The SDN control plane failure emergency method according to claim 6, wherein, when a direct connection failure occurs between the part of the switches and the controller, further comprising: The switch detects through the ECHO data packets sent regularly, and when the ECHO reply data packet is not received within the specified time, it is judged that the switch is disconnected from the controller, and the switch will try to communicate with the controller. The controller re-establishes the TCP connection. 8. SDN control plane fault emergency system according to claim 6, is characterized in that, after described control emergency system enters single data plane self-learning mode, further comprises: Processing the APR packet by adding the entry, and processing the OSPF packet by adding an action, wherein the action includes an entry for a special action; Perform self-study of the table entry on the received data packet. 9. SDN control plane failure emergency method according to claim 8, is characterized in that, described entry self-study is carried out to the described data packet that receives, further comprises: Complete the study of MAC address and port number, and guide the final forwarding port of the data packet; Learn the corresponding relationship between the IP address and the MAC address of the ARP data packet, and guide the hosts in other network segments to request the forwarding of IP data packets between the hosts in the local network segment; Calculate the next-hop switch to each target network segment in the network, and guide the forwarding of IP data packets between hosts in different network segments in the later stage. 10. The SDN control plane failure emergency method according to claim 6, characterized in that, said carrying out said entry self-study to said data packets received, further comprising: The data packets are guided to realize jumping between different network layers.