CN105072037A - Dynamic flow generating method for distributed SDN controllers - Google Patents

Abstract

The method for generating a dynamic flow table oriented to a distributed SDN controller provided by the present invention can effectively reduce the number of flow tables generated by the controller, increase the bandwidth of the management network, reduce the consumption of switch storage resources, and thereby improve the overall performance of the network. The method of the present invention is to regard a network area managed by a controller as a network autonomous domain, and when the flow request received by the controller is a data packet sent to the domain, a shortest path algorithm is used to generate a flow table; If the flow request received by the controller is to send data packets from within the domain to outside the domain, the controller first obtains the global topology views of all networks managed by the distributed controller, generates a global flow table, and In the flow table, the flow table belonging to the local domain is delivered to the switch in the local domain, and the flow table not belonging to the local domain is sent to the controller to which it belongs, and the distribution function is completed by the corresponding controller.

Description

A dynamic flow table generation method for distributed SDN controllers

technical field

The invention relates to a method for generating a dynamic flow table oriented to a distributed SDN controller, belonging to the technical field of computers.

Background technique

In the traditional Internet, the control center and forwarding center are concentrated in a single network device, which makes the existing network control plane increasingly complex. At the same time, the application of new management technologies to the network requires the configuration of existing network devices one by one. , greatly increased the cost, and seriously lacked flexibility and scalability. Therefore, the concept of decoupling the control plane and the forwarding plane is proposed. As long as both ends are connected according to the defined specifications, they can develop independently without being restricted by the other party. This is exactly the essence of SDN technology. Foundation, its early research work includes 4D architecture, RCP, SANE and Ethane, etc. The control plane and the forwarding plane need to be connected according to a unified specification, which is a communication protocol. In 2008, the CleanSlate organization of Stanford University in the United States proposed the OpenFlow protocol, which is currently developed by the ONF (Open Network Foundation) and is the mainstream protocol between the control plane and the forwarding plane. Currently, OpenFlow1.4 has been released. In academia, the current academic research on SDN is very active. According to preliminary statistics, there have been more than 20 academic papers on SDN in top international academic conferences on computer networks such as SIGCOMM, NSDI, OSDI, and CoNEXT. Device design, new data plane design, programming language and interface, testing and debugging, applications (including data center applications, campus network applications, network security applications, etc.).

SDN is facing many unresolved technical problems, among which the most noteworthy centralized SDN controller encounters performance bottlenecks. Due to the uncertainty of network scale, the centralized controller can no longer meet the rapidly changing network requirements. For example, the centralized controller NOX can support 30K requests per second, a 1500 server cluster generates 100K requests per second, and a data center with 100 switches may generate 10000K requests per second. At present, the scale of the backbone network of the operator is 1000 to 2000 devices, and the scale of the data center network with 10000 hosts is about 300 devices. Therefore, centralized controllers are completely unable to meet current network needs. Therefore, with the increase of network scale and business requirements, it is necessary to study the scalability solution of the control plane, that is, the distributed multi-controller solution.

With the expansion of network scale and the increase of SDN upper-layer applications, a large number of flow entries will be generated in the network, which greatly reduces network performance. At present, the number of flow entries in the data center has exceeded 100,000, googleWAN has nearly 143,000; the backbone network of Stanford University has 757,000; Internet2 has 100,000. Flow entries of this magnitude will add a lot of load to switches with limited resources, degrading their forwarding performance. However, in the existing distributed controllers, the shortest path algorithm is used to generate the flow table, and the network managed by the distributed controller is not analyzed, so that the network and the forwarding device will be generated when the flow table is generated and stored. Larger loads degrade overall network performance.

Contents of the invention

1. Purpose:

The main purpose of the present invention is to provide a dynamic flow table generation method for distributed SDN controllers, which can effectively reduce the number of flow table entries generated in the distributed controller, reduce the pressure on the forwarding plane storage flow table, and effectively reduce the transmission The network bandwidth occupied by the flow table reduces the query delay of the switch managed by the controller, improves the forwarding capability, and makes the entire network have higher performance.

2. Technical solution:

In order to realize above-mentioned method, technical scheme of the present invention is such:

When generating the flow table, the solution of the present invention obtains the global topology map through the distributed SDN controller, regards the area within the management scope of a controller instance as an autonomous domain, and generates the flow table according to the ordinary method for the routing rules in the domain; for For routing rules outside the domain, only one flow entry is generated between domains, and data packets are forwarded between domains according to the labeled flow table.

A method for generating a dynamic flow table oriented to a distributed SDN controller, characterized in that the network area managed by a controller is regarded as a network autonomous domain, and when the flow request received by the controller is to send data packets in the domain , then use the shortest path algorithm to generate the flow table; if the flow request received by the controller is to send data packets from within the domain to outside the domain, the controller first obtains all the networks managed by the distributed controller The global topology view, generate a global flow table, send the flow table in the global flow table that belongs to the local domain to the switch in the local domain, and send the flow table that does not belong to the local domain to the controller to which it belongs, and the controller completes it Send function.

When generating the global flow table, different flow table rules are generated according to the location of the switch. The generated flow table is as follows:

Non-exit switches in the intra-domain network: add tag labels to data packets sent outside the domain;

The egress switch in the intra-domain network: forwards the tagged data packet to the intermediate network;

●Switches in the intermediate network: forward the data packets with the tag to the external network;

The ingress switch in the extra-domain network: remove the tag information in the tagged data packet;

● Non-ingress switches in the extra-domain network: forward data packets according to the destination information of the data packets.

Several explanations about the present invention:

(1) SDN: Software-defined networking (Software-defined networking), a network virtualization (Network virtualization) technology, was proposed by Emulex. Using the OpenFlow protocol, the control plane (controlplane) of the router is separated from the data plane (dataplane), and implemented in software. This architecture allows network administrators to re-plan the network with a central control method without changing the hardware device. It provides a new method for controlling network traffic and provides a good platform for core network and application innovation. .

(2) SDN controller: SDN is a new type of network architecture that divides the network into a data plane and a control plane. The data plane is mainly responsible for data forwarding, while the control plane is responsible for formulating corresponding forwarding strategies to guide the data plane. As a part of the SDN network architecture, the control plane is sometimes called "Network Operating System (NOS).

(3) Flow table: the English word corresponds to Flow. In the SDN network architecture, the controller manages the switch by sending flow table information to the switch. The switch can use the flow table to generate a forwarding table. operate.

(4) Autonomous domain: In the present invention, the network range managed by a controller is regarded as an autonomous domain, and each autonomous domain includes a controller, a certain number of switches, and hosts connected behind the switches.

3. Advantages and effects:

The distributed SDN controller-oriented dynamic flow table generation method disclosed by the present invention has the following main advantages:

(1) The dynamic flow table generation method for distributed SDN controllers can make full use of the scenarios of distributed SDN controllers, effectively reduce the flow table item data in the entire network, and generate flow tables according to the global topology. Optimization to achieve the highest network performance with the least flow table. In the distributed SDN controller scenario, there is a large amount of data packet communication between the networks managed by multiple controllers. Therefore, each network can be regarded as an object, and the entire network can be simplified as the communication of multiple objects. In this way, various routing information in the original network can be transformed into communication between two objects. In this way, the original routing information can be greatly simplified, so that when generating the flow table, a small number of flow tables can be used to achieve the original purpose of managing the network.

(2) Reduce the bandwidth occupied by the management network. In the original flow table generation scheme, the management network needs to issue a large number of flow table entries, such as hundreds of thousands of flow table entries in the data center network, which will occupy a large amount of management network bandwidth, increase the network delay, and reduce the network performance. After the flow table is generated using this solution, the flow table items in the management network can be greatly reduced, so that when the flow table is issued, the network bandwidth occupied by the flow table can be effectively reduced, the bandwidth utilization rate of the overall network can be effectively improved, and the network can be improved. overall performance.

(3) Reduce the consumption of switch storage resources. For switches, the storage resources of general switches are relatively limited. When the controller generates too many flow entries, each switch will store a large number of flow tables, wasting the storage resources of the switch. It will also increase the time for querying entries and reduce forwarding performance; after the number of delivered flow tables is reduced, the consumption of flow tables on switch storage resources can be reduced, and at the same time, the time consumption of data packets when querying forwarding tables can be increased, and forwarding efficiency can be significantly improved , so that the network performance can be improved.

Description of drawings

Figure 1 is a structural diagram of a single autonomous domain;

Figure 2 is a structure diagram of dynamic flow table generation for distributed SDN controllers

Fig. 3 is the flowchart of flow table generation;

Detailed ways

In order to make the object, technical solution and advantages of the present invention more clearly, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific examples.

The present invention is a dynamic flow table generation method for distributed SDN controllers. Figure 1 shows the structure of an autonomous domain, which includes a controller, several switches, and many hosts connected to the switches. The controller manages the switches by sending flow tables, so as to achieve the purpose of managing and scheduling the network. The technology proposed by the invention achieves the purpose of managing the network when generating as few flow tables as possible, thereby improving the overall performance of the network.

The technical solution will be described below with the overall structure diagram of the distributed SDN controller. The structural diagram is shown in Figure 2. In the figure, the network managed by the distributed controller is appropriately simplified, and only three autonomous domains are taken as examples to explain the specific process of this solution. As shown in Figure 2, network 1 is the source network, which is managed by controller 1; network 2 is the destination network, which is managed by controller 3; and the intermediate network is the network through which data packets need to be delivered to the destination network, and is controlled by controller 2. manage.

A method for generating a dynamic flow table oriented to a distributed SDN controller is to regard a network area managed by a controller as a network autonomous domain, and when the flow request received by the controller is to send data packets within the domain, then Use the shortest path algorithm to generate the flow table; if the flow request received by the controller is to send data packets from within the domain to outside the domain, the controller first obtains the global topology of all networks managed by the distributed controller view, generate a global flow table, send the flow table in the global flow table that belongs to the local domain to the switch in the local domain, and send the flow table that does not belong to the local domain to the controller to which it belongs, and the distribution function is completed by the subordinate controller .

When generating the global flow table, different flow table rules are generated according to the location of the switch. The generated flow table is as follows:

Non-exit switches in the intra-domain network: add tag labels to data packets sent outside the domain;

The egress switch in the intra-domain network: forwards the tagged data packet to the intermediate network;

●Switches in the intermediate network: forward the data packets with the tag to the external network;

The ingress switch in the extra-domain network: remove the tag information in the tagged data packet;

● Non-ingress switches in the extra-domain network: forward data packets according to the destination information of the data packets.

As shown in Figure 2, when controller 1 receives a flow request from its domain, controller 1 generates a flow table according to the global topology, and sends its own flow table, while the flow table belonging to other domains is sent to other domains. The controller sends the flow table by other controllers. For the autonomous domain managed by controller 1, when the switch matches a data packet sent from network 1 to network 2, the switch will add a tag to the data packet according to the rules of the flow table. This tag is globally unique and determined by the controller. generated by the router; when the data packet is delivered to the intermediate network through the egress switch of network 1, the switch of the intermediate network has a flow table corresponding to the tag, and the switch forwards the data packet to the target network according to the matched tag information; the target network When the ingress switch matches a data packet with the tag, it removes the tag contained in the data packet, and then sends the data packet to the destination host according to the normal flow table. It can be seen from the figure that in the intermediate network, only one flow table is used to complete the data packet forwarding from the entire network 1 to the network 2, while in the existing mechanism, each switch needs dozens or even hundreds of flow tables. flow entries to complete this task.

The present invention makes full use of the characteristics of the distributed SDN controller. The controller that calculates the flow table can obtain the global topology view of the network managed by the distributed controller. When generating the flow table, a network autonomous domain is regarded as an object. The generated flow table takes the autonomous domain as the basic unit. For all data packets sent from network 1 to network 2, add a label in autonomous domain 1; forward according to a flow table in the intermediate network (the flow table is in the unit of a switch, and each switch has a flow table); Remove the label in autonomous domain 2. Thereby effectively reducing the number of flow tables in the intermediate network.

In a network managed by a distributed SDN controller, each controller collects a topology view of the network it manages. Therefore, the entire distributed SDN controller will collect a global topology view of all networks. When calculating the flow table, the controller uses the above algorithm to calculate the flow table of all nodes according to the topology view. After completion, if it is the flow table of this autonomous domain, it will issue it by itself; if it is the flow table of other autonomous domains, then Send the flow table to other controllers, and other controllers complete the delivery function.

The following is a specific flowchart for generating a dynamic flow table for distributed SDN controllers to illustrate this technical solution, as shown in Figure 3:

Step S01: The host in network 1 sends a flowrequest to controller 1, asking how to send the data packet to network 2, and jumps to step S02.

Step S02: Controller 1 judges whether the destination network of the stream request is Network 2; if yes, jump to step S06, otherwise jump to step S03.

Step S03: Determine whether the destination network of the stream request is the current network; if yes, go to step 04, otherwise go to S05.

Step S04: According to the normal process, use the shortest path algorithm to generate the flow table, and the operation is completed.

Step S05: The destination network of the flow request is another network (referring to the network managed by the distributed controller, except the network managed by controllers 1, 2, and 3), and the flow table is generated through the process of other networks, and the operation is completed.

Step S06: Controller 1 requests the global topology data from Zookeeper, and obtains the management and control information of each controller, so as to know which controller each switch is managed by, and skips to step S07.

Step S07: Controller 1 uses the obtained global topology information to calculate the path through which data packets are sent from network 1 to network 2, and generates a global flow table. The rules of the generated flow table are explained in detail below, and skip to step S08.

Step S08: The controller 1 sends the flow tables belonging to the autonomous domain in the global flow table to the switch, and jumps to step S09.

Step S09: Controller 1 sends the flow table not belonging to the autonomous domain to the controller to which it belongs, and then jumps to step S10.

Step S10: After the other controllers receive the flow table sent by controller 1, they send the flow table to the switches they manage, and then jump to step S11.

Step S11: After receiving the flow table sent by the controller, all the switches install the flow table into the switch, and jump to step S12.

Step S12: The switch forwards the data packet according to the rules of the flow table, and the operation is completed so far.

When generating the flow table, different flow table rules will be generated according to the location of the switch. The generated flow table is as follows:

●Non-exit switch in network 1: add tag to the data packets sent to network 2.

●Exit switch in network 1: forward the tagged data packet to the intermediate network.

●Switches in the intermediate network: forward the tagged data packets to network 2.

●Ingress switch in network 2: remove the tag information in the tagged data packet.

● Non-ingress switches in network 2: forward data packets according to the destination information of the data packets.

It can be seen from the above that the main idea of the present invention is to propose a dynamic flow table generation technology for the scenario of a distributed SDN controller, so as to reduce the number of generated flow entries in the entire network, and at the same time, reduce a large number of flow table entries. Table entries consume switch storage resources and network bandwidth, improving the overall performance of the network.

In terms of software, the present invention has no requirements on the operating system, because the Java operating environment is used, and this set of software can run in any environment that supports Java, including Linux systems, Windows systems, MacOS, etc. The recommended Java version is 1.8.0. For the switch in the network, it needs to support the Openflow protocol, the version is 1.3.

Finally, it should be noted that the above embodiments are only used to illustrate and not limit the technical solutions of the present invention, although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: the present invention can still be modified Or an equivalent replacement, any modification or partial replacement without departing from the spirit and scope of the present invention shall fall within the scope of the claims of the present invention.

Claims (2)

1. A dynamic flow table generation method for a distributed SDN controller, characterized in that: The network area managed by a controller is regarded as a network autonomous domain. When the flow request received by the controller is to send data packets in the domain, the shortest path algorithm is used to generate a flow table; if the controller receives When the flow request is sent from the domain to the outside of the domain, the controller first obtains the global topology view of all networks managed by the distributed controller, generates a global flow table, and lists the data packets belonging to the domain in the global flow table The flow table is sent to the switch in the local domain, and the flow table that does not belong to the local domain is sent to the controller to which it belongs, and the distribution function is completed by the controller to which it belongs. 2. the method for generating a dynamic flow table facing a distributed SDN controller according to claim 1, characterized in that: when generating a global flow table, different flow table rules are produced according to the position where the switch belongs, and the generated flow The table is as follows: Non-exit switches in the intra-domain network: add tag labels to data packets sent outside the domain; The egress switch in the intra-domain network: forwards the tagged data packet to the intermediate network; ●Switches in the intermediate network: forward the data packets with the tag to the external network; The ingress switch in the extra-domain network: remove the tag information in the tagged data packet; ● Non-ingress switches in the extra-domain network: forward data packets according to the destination information of the data packets.