WO2014026512A1

WO2014026512A1 - Processing method for applying rapid spanning tree protocol to single ring network topology

Info

Publication number: WO2014026512A1
Application number: PCT/CN2013/078554
Authority: WO
Inventors: 王文龙; 周旭峰; 杨贵; 刘明慧
Original assignee: 南京南瑞继保电气有限公司; 南京南瑞继保工程技术有限公司
Priority date: 2012-08-15
Filing date: 2013-07-01
Publication date: 2014-02-20
Also published as: CN102821050A; CN102821050B

Abstract

Disclosed is a processing method for applying a rapid spanning tree protocol to a single ring network topology, comprising: adding two additional protocol fields in a BPDU: field 1 indicating the protocol type, field 2 indicating topology change information, if a port is invalid, setting the value of the field 2 to be true; otherwise setting it to be false; a switch, after receiving the expanded BPDU, on the one hand, matching the field 1, and when one or more switches do not support or are not configured into a single ring network operating mode, all the other switches on the network automatically operating the standard rapid spanning tree protocol so as to cooperatively work; and on the other hand, judging the value of the field 2, if the value is true, an RSTP state machine sending the BPDU with the true field value so as to notify a counter-side switch of link failure on the network, and each switch conducting rapid fault restoration. The method can shorten the self-healing time of a ring network when a link in a single ring network topology is invalid, and at the same time can be automatically compatible with a standard RSTP, thereby meeting the requirements of various communication services.

Description

A fast spanning tree protocol applied to a single ring network topology processing method

Technical field

The present invention relates to communication network redundancy technologies, and in particular, to a fast spanning tree protocol applied to a single ring network topology. Background technique

Spanning Tree Protocol (STP), Rapid Spanning Tree Protocol (RSTP), and Multiple Spanning Tree Protocol (MSTP) are topological management protocols that apply to Layer 2 communication networks. The Layer 2 switch runs the spanning tree protocol and sends the Bridge Protocol Data Unit (BPDU) to each other. The main information in the BPDU includes the STP version, the BPDU type, the identifier field, the root bridge ID, the root path cost, and the specified bridge ID. The port ID and the time value are specified. The root bridge ID, the root path cost, the specified bridge ID, and the specified port ID form the message priority vector, which is the main basis for RSTP comparison and calculation. The switch obtains the information in the BPDU and uses the spanning tree algorithm (SPA) to tailor the network topology to eliminate loops and form redundant paths. When a network link fails, STP can use redundant paths to restore normal communication services. Running the spanning tree protocol through the Layer 2 switch can eliminate loops in the communication network to avoid network storms and implement redundant backup of links.

The Rapid Spanning Tree Protocol is an improvement to the spanning tree protocol. By introducing the Proposal/Agreement mechanism (referred to as the P/A mechanism), the neighboring switches can quickly negotiate to tailor the network topology. Compared with the spanning tree protocol, the stable network topology is formed. Significantly improved, it can achieve a recovery speed of about 1-2 seconds.

For the sake of generality, the Rapid Spanning Tree Protocol applies to any network topology in accordance with the spanning tree algorithm and the P/A Negotiation Machine. When the switch receives the BPDU, it first collects the useful information in the BPDU, and then performs the calculation according to the spanning tree algorithm to determine the port role, including the root port, the designated port, the alternate port, and the alternate port. (Backup Port) and the status of the port, including the block. The state in which the port does not forward packets and does not learn the address is called Discarding, which is equivalent to the Block state in STP. The state is called blocking state, learning, forwarding, and finally sending BPDUs to the opposite side to complete the negotiation process. The process grows between switches on the network, such as trees, from tree roots to leaves, and sequentially between the two, thus forming a tree-like topology with no loops.

The implementation of the fast spanning tree algorithm in the switch is accomplished through a set of state machines that work together. There are a total of 10 state machines. Each state machine contains several "states", operations within each "state", and conditions for entering "state". When in a "state", the program performs the predetermined operation under the "state". After the execution is completed, Check if there are other "states" that meet the entry conditions, and if so, move to the "state", otherwise keep the current "state" and wait for the next condition check. The state machines interact with each other through shared variables to affect the "state" migration conditions.

When the fast spanning tree algorithm is implemented by software, multiple state machines can only be executed in a certain order because of the sequence execution of the program. When a conditional change affects a state machine, the program must check the state machine's migration condition from the beginning. Most of the time, this check process introduces a certain processing delay. On the other hand, when there are multiple state machines to process, once a state machine has a time-consuming operation, it will affect the execution of all subsequent state machines.

The ring network is a widely used network topology with clear structure and redundant backup function to provide single point link failure protection. The ring network has many forms, such as a single ring network, a double ring network, a tangent ring, and the like. In actual engineering, including communication and other industrial fields, single-loop network topology has a wide range of applications, taking into account factors such as fault protection capability and cost. The characteristics are: The switches on the ring each use two ports to connect with adjacent switches. The other ports are connected to the non-spanning tree port on the terminal node or other switch (that is, the port does not belong to any spanning tree). The topology has one and only one redundant link.

RSTP is applied to a single ring network. When a link fails, RSTP ensures that no transient loop is formed. The switch must first block the spanning tree port and then send BPDUs to neighboring switches. When the neighboring switch meets the conditions and replies with the packet. After the notification or forwarding delay (Forwarding Delay), the local blocked port can be converted into the learning and forwarding state in turn, thus extending to the entire network. Under the most severe conditions, half of the switches on the ring need to negotiate and repeatedly block and open. The port can recover the communication service, and the user data stream will be interrupted during the process.

The above process ensures that loops are not formed during topology reconstruction, resulting in network storms. However, considering the characteristics of a single ring network: There is only one redundant link, then after RSTP is stabilized, if any link fails, it is impossible to form a loop. The negotiation process of RSTP in the case of single-ring network link failure is complete but not fast enough.

When the topology changes, RSTP sends a BPDU with the Topology Change flag set. When other switches receive the topology change message, the address update operation is required. According to statistics, the address update operation takes a certain amount of time, usually several milliseconds or even ten milliseconds per station. From the above analysis of the fast spanning tree state machine implementation, this delay will affect the execution of the entire state machine sequence, significantly affecting the link recovery time. For applications such as TCP and UDP, the recovery time of 50ms can meet the requirements. For the industrial field, it is basically real-time communication, and the recovery time is very high. The real-time communication is mainly based on multicast messages (which are broadcasted in the group) or broadcast messages, and does not need to update the address table to restore communication.

It can be seen that when the single ring network topology is managed by using the RSTP protocol, the integrity and security of the network are guaranteed, but it cannot be According to the ring network topology and the characteristics of different types of packets, the ring network has a longer self-healing time and cannot meet the requirements of fast self-healing of the ring network. Summary of the invention

The object of the present invention is to provide a method for processing a fast spanning tree protocol applied to a single ring network topology, which can shorten the ring network self-healing time when the link fails in a single ring network topology, and can be automatically compatible with the standard RSTP. Requirements for various communication services.

In order to achieve the above object, the solution of the present invention is:

A fast spanning tree protocol is applied to a single ring network topology processing method. The BPDU is extended, and two additional protocol fields are added: field 1 indicates the protocol type, and field 2 indicates the topology change information. If the port fails, the value of the field 2 is set. True, otherwise false;

The above topology change identifier is Boolean, and the value is true or false. When there is a port failure, it is set to true, otherwise it is false; the topology change timer is 1-2 seconds.

After receiving the extended BPDU, the switch determines whether the neighboring switch works in the single ring network by matching the field 1 to further determine the working mode of the switch on the entire ring network; when one or more switches do not support Or not configured as a single ring network working mode, all other switches on the network automatically run the standard fast spanning tree protocol to work together; on the other hand, determine the value of field two, if true, the RSTP state machine sends the field value The true BPDU informs the peer switch that there is a link failure on the network, and each switch performs a fast recovery.

The method for quickly recovering the above faults includes the following contents:

(1) Add two variables to each ring network port: topology change identifier and topology change timer;

(2) Detecting the port status. When the port connection is restored, it is necessary to ensure that no instantaneous loop occurs, and the standard RSTP state machine is used for processing;

When the port connection fails, the topology change timer of the port is started immediately, and the fault type is distinguished according to the port role: Type 1, the alternate port fails; Type 2, the non-alternative port (ie, the root port or the designated port) fails, and the The other ring port of the switch where the port is located is an alternate port; type 3, the non-alternative port is invalid, and the other ring port of the switch where the port is located is not an alternate port;

When a type 1 fails, the fault does not cause a topology change because it is a redundant link failure.

When the type 2 is faulty, the RSTP state machine quickly puts the alternate port of the local machine into the forwarding state, and the fault is instantaneously restored. When the type 3 fails, the topology change identifier of the other ringed port that is in the connected state is set to true, and the connection port is The RSTP state machine variable newlnf 0 is true, and the topology change timer of the connection port is started at the same time. The port state is converted to Block;

For the definition of the above newlnfo variable, see IEEE 802. ID-2004 17. 19. 16, Boolean, the value is true or false. When it is true, it indicates that new information needs to be sent through the BPDU, and the port Transmit state machine (IEEE 802. 1D-2004 17. 26) will be started.

(3) When the topology change information field is set in the BPDU received by the switch, indicating that the link in the network is invalid, the topology change timer of the port is started immediately. At the same time, the state machine checks. If the port is an alternate port, Immediately convert to the forwarding state; otherwise check another ring port, if it is an alternate port, convert to the forwarding state; if there is no alternate port, set the topology change identifier and newlnfo flag of the other port;

(4) When the topology change timer of the port is not 0: The RSTP program does not check the status of the port, and does not process the topology change information field set in the BPDU. The RSTP port state transition state machine (IEEE) 802. 1D-2004 17. 30) The state transition of this port is not performed.

A fast spanning tree single ring network processing method further includes:

(1) When the RSTP program executes each state machine, the port transmission state machine is first executed, so that the port transmission state machine can detect the change of the port variable newlnfo at the first time. When the variable is true, the txRstp() function uses the The topology change of the port identifies the BPDU additional field 2, topology change information, and immediately sends a BPDU to notify the peer switch of the topology change event.

For the definition of the above txRstp O function, see IEEE 802. 1D-2004 17. 21. 20, the function is to fill the BPDU with each variable of the port and send a fast spanning tree BPDU.

(2) The Topology Change state machine (IEEE 802. 1D-2004 17. 31) detects a topology change or receives a topology change flag (Topology Change flag, IEEE 802. 1D-2004 9. 3. 3) When setting the BPDU, set the port variable fdbFlush (IEEE 802. ID- 2004 17. 19. 7) to clear the port address table, so that the switch can learn the MAC address again, that is, the address update, to restore the communication service. The address update process is separated from the internal operations of the RSTP topology change state machine and a separate address updater is used to detect the fdbFlush state. When fdbFlush is set, the address update program immediately resets fdbFlush, but after a certain delay, the address update operation is performed. The length of the delay depends on the size of the network, typically 20ms.

After adopting the above scheme, the invention has the following characteristics:

(1) Add field transmission topology change information to the BPDU, so that the RSTP program can directly obtain the network failure information through the field, and can respond without complicated calculation of the state machine;

(2) Make full use of the characteristics that the single ring network has one and only one redundant link. When the fault is detected, the alternate port is directly changed to the forwarding state, and the network communication service is quickly restored. (3) First, the RSTP port is sent to the state machine. When the port topology change identifier and the newlnfo flag are set, the RSTP state machine sends a BPDU to the opposite switch at the first time to quickly propagate the fault information to the entire network.

(4) Using an independent address update program for address update operation, avoiding the time-consuming process of the address update process affecting the calculation process of the state machine, and satisfying the requirements of various communication services for the network self-healing time, and simultaneously speeding up The convergence speed of RSTP in the case of link failure and link recovery. DRAWINGS

1 is a schematic diagram of an extended fast spanning tree BPDU format;

2 is a schematic diagram of a single ring network topology;

Figure 3 is a schematic diagram of the flow of the RSTP single ring network processing method. detailed description

In order to make the technical solutions and features of the present invention more clear, the present invention will be further described below in conjunction with the accompanying drawings. In the embodiment of the present invention, the Rapid Spanning Tree BPDU (RST BPDU) is extended to transmit topology change information. To be compatible with standard BPDUs, to achieve interoperability between different switches, without changing the existing fields, meanings, and locations of the RST BPDUs, only add an additional protocol field at the end of the RST BPDU, the length is 1 byte, The 7-bit (most significant bit) indicates the topology change identifier, called RT (:, the remaining bits (6-0 bits) indicate the protocol type, and the RST BPDU format and meaning of the RTypc extension are shown in Figure 1.

In the embodiment of the present invention, the fast spanning tree single ring network processing mode opening switch is set to a Boolean variable RRING, and the variable is used to control whether the fast spanning tree single ring network processing method is enabled. When 匪G is false, the standard RSTP protocol is run; when RRING is true: The port receives the BPDU, first determines the BPDU protocol version identifier, and if it is the fast spanning tree protocol version, it determines whether the RType of the additional protocol field is equal to the specified type, and satisfies the above two The RTC field is processed when the condition is met. Otherwise, considering the interoperability with switches of different manufacturers, the switch should automatically set RRING to false and compatible with the standard RSTP protocol.

RSTP calculates the topology of a single ring network by selecting a bridge ID (consisting of 16-bit bridge priority and 48-bit MAC address). The smallest switch is used as the root bridge, and the other switches are designated bridges. According to the principle that the root path cost is the smallest (the distance to the root bridge is the shortest), the topology extends from the root bridge to the designated bridges on both sides, and the root path cost increases one by one. The P/A mechanism between the switches ensures that no temporary loops occur. The number of switches on the ring network is N, and the path cost of each switch is 1. The maximum path cost of the switch on the ring is (N/2). The ring is disconnected by the alternate port in the blocked state at the switch where the root path cost is equal to the maximum. The root path cost of the switch is the most The small port is the root port; the port that connects to the downstream network and acts as the path of the downstream network to the root bridge is the designated port; the alternate port is the alternate port of the root port, and when the root port is unavailable, the alternate port can be converted to the root port . After the topology is stable, one or only one of the alternate ports in the ring network is blocked. The other ports are root ports or designated ports and are in the forwarding state.

A single ring network consisting of four switches is shown in Figure 2. According to the above method, a stable network topology is formed: swl is the root bridge, and sw2-4 is the designated bridge; the p2 port of sw4 is the alternate port, which is in the blocking state, and the other port is the root port or the designated port, and is in the forwarding state; For redundant links, user traffic is not forwarded, and other links are links used for normal communication.

The following is an analysis of the RSTP processing method when each link fails.

When link C fails, the sw4-p2 port role is an alternate port, and link C is a redundant link, which is logically disabled. The RSTP state machine can instantaneously converge without affecting the normal operation of other links.

When link B fails, the other ring port p2 of sw4 acts as an alternate port and can be restored to the root port immediately. The RSTP state machine can also converge instantaneously.

When link A or C fails, the RSTP state machine must recalculate the topology. Since RSTP treats the ring network as a normal network, there is no significant difference in self-healing time from the general network.

The fast spanning tree single ring network processing method takes into account the characteristics of the single ring network. The following steps are used to shorten the network self-healing time.

In the embodiment of the present invention, two variables are added to each ring network port: a topology change identifier, which is called PT (:, the value is true or false; the topology change timer is called PTT. For convenience of description, two switches Ring ports are called Α,

Β port.

The following takes the failure of the port or the receipt of the BPDU as an example to explain the processing. The processing of the B port is completely consistent. The detailed process is shown in Figure 3 and explained as follows:

301: Initialize the PTC identifier and PTT timer of the A and B ports of the switch. The PTC flag is initialized to false and the PTT timer is initialized to 0.

302: When the PTT timer is 0, it detects whether the status of the ring network port changes.

303: The port recovery is detected. Considering that there is a complete and stable topology, and the port is restored in the single ring network structure, the temporary topology is incomplete due to unsuccessful P/A negotiation. Regardless of the port role that the recovery port will be converted, RSTP can negotiate successfully in a short period of time, and the topology converges. Therefore, no special processing is required in this case.

In the ring network structure, there is a possibility that the P/A negotiation is unsuccessful. The port needs to go through two forwarding delays. Can enter the forwarding state. The analysis shows that the location of the port is only on the opposite side of the alternate port. The link connecting the two ports is a redundant link. Whether the port is in the forwarding state does not affect the normal operation of the network.

304, 310: If the port is detected to be invalid, or the BPDU with the RTC flag is set, the PTT timer of the port is started immediately. The timing is recommended to be 1-2 seconds.

305: If the A port fails and is an alternate port, the state machine itself can converge quickly without special handling.

306: When the A port is not an alternate port, if the B port is an alternate port, the B port is converted to the forwarding state, and the B port PTT timer is started, and then the RSTP state machine is executed.

307: When the B port is not an alternate port, set the PTC and newlnfo flags of the B port to start the PTT timer. The A port is switched to the blocking state, and finally the RSTP state machine is executed.

308: The port B receives the BPDU. When the PTT timer of the port is 0, the RTC identifier field of the packet is detected.

309: If the RTC of the BPDU received by the A port is not set, it indicates that there is no topology change event, and the RSTP state machine is executed for normal processing.

310: Same as 304.

311: When the received RTC flag of the BPDU is set, if the A port is an alternate port, the port transits to the forwarding state, and then the RSTP state machine is executed. Otherwise go to 306 to execute.

320: The txRstp O function fills the RTC field of the port BPDU according to the PTC identifier of each port, that is, the value of the RTC field of the BPDU is equal to the value of the port PTC, and the RType field, and then sends a BPDU.

It can be seen from Figure 3 that 301-307 constitutes the port state detection process to detect in real time whether the state of the ring network port has changed. 308-311, 306, and 307 form a receiving process of the BPDU. Both of the above processes are triggered by events, and therefore, the response speed of the RSTP program to the event significantly affects the implementation effect of the present invention.

In order to improve the response speed of the program to the event, the embodiment of the present invention adopts a real-time operating system and sets the RSTP task to a higher priority to ensure good response speed and sufficient execution time. Both the port status change event and the BPDU receive event use the interrupt mode to notify the RSTP program to improve the overall performance of the system.

It can also be seen from Figure 3 that the key process for rapid self-healing of the ring network is to quickly find an alternate port and convert it to a forwarding state. The alternate port has two locations for the switch where the fault is located: local and other switches. If the RSTP program is considered to have a good event response speed, the process can be completed instantaneously when the alternate port is a local port; if the alternate port is on another switch, the alternate port must be found by the BPDU propagation.

The length of the fast spanning tree BPDU is 36 bytes, and the additional protocol field is 1 byte (see Figure 1). When an Ethernet packet is formed, the BPDU header adds 6 bytes of destination address, 6 bytes of source address, and 2 characters of length. Section, LLC field 3 bytes, total 54 bytes, fill the minimum Ethernet frame, a total of 64 bytes, network propagation delay is very small (100Mbps connection, The transmission time is 6.72 microseconds. When the network size is not too large, the delay of the part is not considered relative to other processing delays; RSTP runs on the second layer, and the protocol delay is very small; therefore, the transmission delay of the BPDU mainly comes from The delay calculated by the state machine of RSTP itself.

In the embodiment of the present invention, the port sending state machine is placed in the first place of the RSTP state machine list, and once the port newlnfo flag is set, the BPDU is sent immediately. In this way, the delay from the RSTP state machine calculation during BPDU propagation can be minimized. According to step 307 of Figure 3, the port PTC identifier is set at the same time as the newlnfo flag, so that the PTC identifier can be sent with the BPDU at the first time, and the original or forwarding speed is improved, thereby effectively reducing the ring network self-healing time.

The solution provided by the embodiment of the present invention to prevent the RSTP state machine from performing unnecessary port state transitions in the calculation process is:

When RRING is false, or the port PTT timer is 0, the actual operation of the port state transition can be performed, but the above conditions should not affect the state transition process of the RSTP port state transition state machine. In this way, after the topology change event is detected and the PTT timer expires, the RSTP state machine will not perform the actual port state transition operation. Properly configuring the PTT timer value can avoid the RSTP topology recalculation process. This saves the port state transition time and speeds up the link recovery without affecting the normal operation of the RSTP state machine.

The embodiment of the present invention further delays the RSTP address update operation according to the different requirements of different communication services for the ring network self-healing time and the characteristics of different message transmission modes, and accelerates the calculation process of the RSTP state machine. On the other hand, the communication services based on different message types meet the requirements for different self-healing times of the ring network.

When the port's fdbFlush flag is set, the address update operation is pushed into the delay task processing queue, fdbFlush is immediately cleared, and the delay time is 20 milliseconds. When the delay is full, the address update operation is performed. This separates the RSTP calculation process and the address update operation process, and avoids the address update operation taking up the calculation time of the RSTP state machine. During the delay period, the communication physical link has been quickly restored, and the broadcast and multicast packets are forwarded normally. The communication services based on these two types of packets only lose a small amount of packets. After the delay, the address update operation is performed. In 50ms, services such as TCP can be restored to meet the needs of other applications.

The embodiment of the invention provides a fast spanning tree single ring network processing method, which can significantly shorten the ring network self-healing time when the link fails in the single ring network topology, and can be automatically compatible with the standard RSTP to meet the requirements of various communication services. .

The above embodiments are only for explaining the technical idea of the present invention, and the scope of protection of the present invention is not limited thereto. Any changes made based on the technical solutions according to the technical idea of the present invention fall within the protection scope of the present invention. Inside.

Claims

claims

1. A processing method for rapid spanning tree protocol applied to single ring network topology, which is characterized by: extending BPDU and adding two additional protocol fields: field one represents the protocol type, and field two represents topology change information. If the port fails , the value of field two is set to true, otherwise it is false;

After the switch receives the extended BPDU, on the one hand, it determines whether the adjacent switch is working in the single-ring network processing mode through matching field 1. When one or more switches do not support or are not configured for the single-ring network working mode, the network All other switches on the network automatically run the standard Rapid Spanning Tree Protocol to work together; on the other hand, determine the value of field two. If it is true, the RSTP state machine sends a BPDU with a true value in this field to inform the opposite switch that there is a link on the network. If the path fails, each switch will quickly recover from the failure.

2. A processing method for rapid spanning tree protocol applied to single-ring network topology as claimed in claim 1, characterized in that: the topology change identifier is a Boolean type, and the value is true or false, when a port fails , set to true, otherwise false; the topology change timer lasts for 1-2 seconds.

3. A processing method for rapid spanning tree protocol applied to single ring network topology as claimed in claim 1, characterized in that: the method for rapid fault recovery includes the following:

(1) Add two variables to each ring port: topology change identifier and topology change timer;

(2) Detect the port status. When the port connection is restored, use the standard RSTP state machine for processing; when the port connection fails, immediately start the topology change timer of this port and distinguish the fault type according to the port role: Type 1, replacement Port failure; Type 2, the non-replacement port fails, and the other ring network port of the switch where the port is located is a replacement port; Type 3, the non-replacement port fails, and the other ring network port of the switch where the port is located is not a replacement port;

When a Type 1 failure occurs, since the redundant link fails, the failure will not cause topology changes;

When a Type 2 fault occurs, the RSTP state machine quickly puts the local alternative port into the forwarding state, and the fault recovers instantly; when a Type 3 fault occurs, the topology change flag of another ring network port in the connected state is set to true, and the connection port's topology change flag is set to true. The RSTP state machine variable newlnfo is true, and the topology change timer of the connection port is started at the same time, and the failed port status is converted to Block;

(3) When the switch receives the BPDU, the topology change information field is set, indicating that a link in the network fails, and immediately starts the topology change timer of the port. At the same time, the state machine checks. If this port is a replacement port, Immediately transition to the forwarding state; otherwise, check another ring network port, and if it is a substitute port, transition to the forwarding state; if there is no substitute port, set the topology change flag and newlnfo flag of the other port;

(4) When the topology change timer of the port is not 0: The RSTP program does not check the status of the port and does not process it anymore. Topology change information field set in BPDU; RSTP port state transition state machine does not perform state transition of this port.

4. The processing method of rapid spanning tree protocol applied to single ring network topology as claimed in claim 1, characterized in that: the processing method also includes the following:

When the RSTP program executes each state machine, it first executes the port sending state machine, so that the port sending state machine detects the change of the port variable newlnfo at the first time. When this variable is true, the txRstp () function uses the topology change identifier of the port. Fill in the BPDU additional field two, and immediately send the BPDU to notify the opposite switch of the topology change event; separate the address update operation from the internal operation of the topology change state machine, and use a separate address update program to detect the fdbFlush identification status of the port. When fdbFlush If set, the address update program immediately resets fdbFlush and delays the address update operation.