CN104270286B - A kind of SDN network node failure localization method - Google Patents

Abstract

In order to be restored automatically when solving the problems, such as to occur in SDN network lost contact failure, the present invention proposes a kind of method by nodal plane to SDN network lost contact fault location, is restored according to the result of positioning.Include the following steps：When keep-alive message disappears, node judges that lost contact failure occurs；If neighbor node state is normal, by controller starter node alternate algorithm, alternative command is issued to substitute node.If neighbor node also lost contact, continue to send inquiry message to other nodes, to each lost contact node within the scope of the lost contact, alternative command is issued from controller to substitute node until determining lost contact range.If the service condition lost contact of lost contact node, migrates existing business.If the whole network node whole lost contact, it is judged as that controller damages, starts controller restoration methods.This programme realizes lost contact Fault Locating Method by node layer and takes corresponding recovery policy, has the characteristics that quick.

Description

A SDN network node fault location method

technical field

The present invention relates to the technical field of SDN network, in particular to a fault location method performed by a node plane and a corresponding recovery method selected when a disconnection fault occurs between a controller and a node.

Background technique

In recent years, SDN network has become a research hotspot, and its own characteristics of separation of control and forwarding have brought new development directions and research hotspots for network development and research.

The SDN network is mainly composed of a control layer and a node layer. The control layer is responsible for resource and business management in the entire network, while the node layer constitutes the network topology and carries the business transmission in the network. The two communicate through the Openflow protocol.

The disconnection fault involved in the present invention refers to the inability to perform normal keep-alive message sending and receiving and Openflow protocol communication between the controller and the node. In an SDN network, when a disconnection fault occurs, how to locate the location of the disconnection fault and then adopt an appropriate recovery strategy will affect the effect of network fault recovery.

In the above SDN network architecture, a disconnection failure may occur at the node plane, at the control plane, or at the communication network between the two. How to use the characteristics and architecture of the SDN network to locate faults as soon as possible through the network node plane itself, and thus adopt different recovery strategies, has become an important factor affecting the robustness of the SDN network.

Contents of the invention

In order to solve the problem of automatic recovery when a disconnection fault occurs in the SDN network, the present invention proposes a method for locating the disconnection fault of the SDN network through a node plane, and the SDN network can recover according to the positioning result.

The SDN network node fault location method of the present invention comprises the following steps:

The SDN network node judges whether the node is disconnected through the keep-alive message between the node and the controller. When the keep-alive message disappears, it is judged that a disconnection fault occurs;

The lost node sends an inquiry message to the neighbor node;

If the state of the neighbor node is normal, it is determined that there is a network fault between the lost node and the controller, or the lost node itself fails; the lost node sends a replacement request to the controller through a normal neighbor node ; The controller starts the node replacement algorithm, and sends replacement instructions to the replacement node.

If the neighbor node loses connection, the neighbor node continues to send inquiry messages to other nodes until the range of loss of connection is determined; the lost connection node sends a replacement request to the controller through a normal neighbor node; for each of the For lost nodes, the controller starts the node replacement algorithm and sends replacement instructions to the replacement nodes.

Further, the disconnected node inquires its own existing business status; if the business status of the disconnected node is abnormal, the controller further starts a business migration algorithm to migrate the existing business of the disconnected node.

Further, when determining the range of loss of connection, if all nodes in the entire network lose connection with the controller, it is determined that the controller is damaged, and a traditional controller recovery method is started.

This solution makes full use of the architecture characteristics and network communication characteristics of SDN itself. Starting from the node plane, without introducing other facilities, the SDN network itself realizes the location method for disconnection faults, so that failures can occur in the SDN network. When a connection fault occurs, locate the fault as soon as possible and adopt a corresponding recovery strategy, which has the characteristics of low consumption and low delay.

Description of drawings

Figure 1 is an architecture diagram of a fault location method based on a node plane.

Fig. 2 is an embodiment of a fault location method based on a node plane.

Detailed ways

Figure 1 is an architecture diagram of a fault location method based on a node plane. The figure is an example of an SDN network of a single domain model, which consists of a controller and 6 nodes. The six nodes are connected to each other to form the "node plane", and the thin black solid lines between the nodes represent the interconnection between the nodes, thus forming the topology in the domain. The dotted line between the controller and the node indicates that the keep-alive message is passed between the two. A dotted line with a cross indicates that the keep-alive message between the controller and the node fails, and node 6 becomes a disconnected node at this time. The arrow sent by node 6 pointing to its neighbor node indicates that the node sends an inquiry message to the neighbor after realizing that it is in a disconnected state. The arrow pointing from the controller to node 3 indicates that the controller determines that node 3 is the replacement node of node 6 after calculating through the node replacement algorithm, and issues a replacement instruction to the replacement node. The dotted line passing through node 2, node 6, and node 4 represents a service path before node 6 loses connection, and the service path is maintained when node 6 is only lost but the service state is normal. If node 6 is both disconnected and has a service failure, the service path needs to be restored. For example, the thick solid line passing through node 2, node 3, and node 4 represents the restoration service path after joint calculation by the node replacement algorithm and the service migration algorithm of the controller.

Fig. 2 is an embodiment of the fault location method of the present invention, and the whole process includes the following steps:

101: Fault judgment, judging whether the node itself is in a normal state or disconnected state according to the connection between the node and the control plane. Under normal circumstances, there is a keep-alive message between the node and the control plane. At the same time, the node can take corresponding actions on the command of the control plane. A disconnection fault as defined in the present invention occurs in the network.

201: The disconnected node sends an inquiry message to the neighbor node.

202: Determine whether the state of the neighbor node is normal.

203: If the status of the neighbor node is normal (not disconnected), it is judged that there is a network failure between the disconnected node and the controller, or the node itself fails (but the controller is operating normally), and the disconnected node passes through the neighbor node Initiate a reconnection, making an alternative request to the controller. Go to step 301;

204: If the neighbor nodes are also out of contact, continue to inquire outside until the range of outage is determined, and go to step 401.

301: The disconnected node inquires its own existing business status, and replies the result to the controller through the neighbor node,

302: Judging whether the business status is normal, divided into the following two situations:

Situation 1: The existing business status of itself is normal, that is, the lost node can no longer carry new business, but it has no impact on the existing business passing through the lost node. Go to step 303 at this point. Situation 2: The existing business status of itself is abnormal, that is, the disconnected node can neither bear the existing business passing through itself nor continue to bear new business. Now go to step 304

303: The controller internally marks this node as a disconnected node. If no new business arrives at this time, the controller does not need other actions for the time being; when a new business passing through this node arrives, the controller triggers the node replacement algorithm, After calculation, a replacement command is issued to the replacement node, and the replacement node carries the new business.

304: The controller starts the node replacement algorithm and the service migration algorithm, and issues a replacement instruction and performs the existing service migration of the disconnected node.

401: According to the number of lost nodes in the network, determine whether all nodes are lost. The scope of disconnection is divided into partial disconnection and complete disconnection:

Case 1: Partially lost connection. If it is judged that some of them are out of contact, go to step 301. When there are nodes in the network that can communicate with the control plane normally, it is a partial loss of connection. At this time, it is determined that the normal operation of the control plane is caused by the failure of some nodes of the node plane. At this time, each disconnected node performs steps 301-304 in order to restore network services to the maximum extent.

Case 2: All connections are lost. Go to step 402.

402: When the controller is damaged and the global domain is lost, start the traditional controller recovery method. The traditional controller recovery method is triggered and executed by the control plane or the network manager, and starts the computer as the standby controller.

After the above steps 303, 304 or 402, the network is restored.

The exception in step 402 is that the controller is not damaged, and the loss of all nodes indicates that all nodes are damaged or the basic hardware of the network is damaged so that the entire network is out of control and cannot be restored through the algorithm.

Claims (3)

1. a kind of SDN network node failure localization method, which is characterized in that include the following steps：

SDN network node by the keep-alive message between this node and controller judge this node whether lost contact, work as keep-alive message When disappearance, it is judged as that lost contact failure occurs；

Lost contact node sends inquiry message to neighbor node；

If the neighbor node state is normal, it is determined as between the lost contact node and controller that there are network failure or institutes State lost contact node itself fail；

If also lost contact, the neighbor node continue to send inquiry message to other nodes, be lost until determining the neighbor node Join range；

Lost contact range is divided into two kinds of situations of part lost contact and whole lost contacts, if judgment part lost contact, the lost contact node passes through Normal neighbor node sends out replacement request to controller；

Lost contact querying node itself existing business state, and result is restored by neighbor node to controller；

To each lost contact node,

Situation one：Itself existing business state is normal, when the new business of the approach node arrives, is replaced by controller triggering node For algorithm, alternative command is issued to substitute node；

Situation two：Itself existing business loss of form is issued replacement to substitute node and is referred to by controller starter node alternate algorithm It enables.

2. SDN network node failure localization method as described in claim 1, which is characterized in that further comprising the steps of：

If the service condition of the lost contact node is not normal, further start business migration algorithm by controller, to the lost contact The existing business of node is migrated.

3. SDN network node failure localization method as claimed in claim 1 or 2, which is characterized in that further comprising the steps of：

When determining lost contact range, it is found that the whole network node all with controller lost contact, is judged as that controller damages, starts traditional Controller restoration methods；

Traditional controller restoration methods, are triggered and are executed by control plane or webmaster, are started and are used as spare control The computer of device.