JP2004282409A - Information transfer system, and logical topology construction method for information transfer system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an information transfer system capable of preventing the abnormality occurrence and stall of the information transfer system. <P>SOLUTION: A master node periodically transmits arbitration RPDU (ring protocol data unit) in order to show that a node is alive and in a good condition. Each slave node always monitors an upper node, discards current topology information when the slave node does not receive the arbitration RPDU for a fixed period or more, returns to the master node and starts arbitration for itself. When the master node fails, this enables all slave nodes to recognize that the master node fails after the lapse of a fixed time, and when a slave node fails, this enables all slaves existing at the downstream of the slave to recognize that the slave fails to be able to restarts logical topology construction work, preventing node down. In addition, a countermeasure against message loss, cache clearance, address learning inhibition, the prevention of a message infinite circulation and a spontaneous blocking are included. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to an information transfer system between a number of frame transfer devices such as a switching hub, and in particular, when the frame transfer devices are connected in a ring, somewhere is logically cut off so that the number of frames is infinite. And a mechanism for preventing the data from being continued to be transferred.
[0002]
[Prior art]
In this type of information transfer system, an information transfer route is provided with a mechanism for logically cutting unnecessary paths (links) from a topology that physically includes loops and constructing a logical tree structure without loops. . Further, a mechanism is provided which can reconstruct the tree structure when a physical or logical disconnection (breakage) occurs between nodes constituting the tree structure.
[0003]
As a mechanism capable of constructing and reconstructing the tree structure, there is a spanning tree protocol (STP) (for example, see Non-Patent Document 1).
[0004]
The STP algorithm is roughly divided into two. One is processing of topology information for forming a tree structure, and the other is processing for supporting refresh of a station cache (change notification mechanism).
[0005]
In order to construct a tree structure, one of the nodes is determined as a root, topology information is exchanged between adjacent nodes for each of the other nodes, and a node whose port (connection end) is closer to the root is determined as “ A process of determining a parent (root port) and a node farther from the root as a "child (representative port)" is executed at all nodes, and topology information of the parent-child relationship is spread throughout the system to construct a tree structure. .
[0006]
In addition, each node does not transmit the received topology information as it is, but is used for updating its own topology information, and transmits the updated information at the time of transmitting the next topology information, so that new topology information is transmitted. Propagating indirectly. The topology information is stored in a BPDU (Bridge Protocol Data Unit).
[0007]
The change notification mechanism is used for reconstructing the tree structure. In this change notification processing, when a topology change is detected, the route is notified.- The root notifies all nodes of the change.- While the topology change notification is issued, the aging time of the station cache is shortened. become.
[0008]
There are the following problems in port state control and topology restoration by the Spanning Tree Protocol (STP) described above.
[0009]
(1) Long time is required to construct a logical topology
When STP is used, it takes several tens of seconds to construct a logical topology. FIG. 7 shows a process until the STP bridge transitions the port to the transfer enabled state (forwarding state).
[0010]
In this process, adjacent bridges exchange topology information with each other through transmission / reception of a setting BPDU, determine the hierarchical relationship between the bridges by comparing their priorities, and finally determine a port role.
[0011]
Here, the port role is a logical role of the port. If the port is a root port or a representative port, the port can transition to the forwarding state, and if the port is an auxiliary port, the port must maintain the blocking state. However, it is not always possible to transition to the forwarding state immediately after being selected as the root port or the representative port, and it takes a long time as described below.
[0012]
In FIG. 7, when the port P receives the setting BPDU, if it is selected as the root port or the representative port by the evaluation of the topology information, the port P transits to the listening state. The port P in the listening state has a transfer delay time (the time required for information transmitted from a certain bridge to spread throughout the system) T _fwd After the elapse, the state is changed to the learning state. Next, the port P in the learning state outputs the transfer delay time T. _fwd After the elapse, the state is changed to the forwarding state.
[0013]
Therefore, the port cannot transfer a general frame until twice the transfer delay time has elapsed since the port role was determined.
[0014]
(2) It takes more time to restore the logical topology when a disconnection occurs.
[0015]
If a disconnection occurs, a new logical topology must be constructed to bypass the disconnection (if possible). In order to quickly construct a logical topology, it is necessary to announce the disconnection to the entire system.
[0016]
In the STP, there is a TCN message as a means for reporting the entire system, but this is used only for clearing the station cache and has no meaning as a trigger for updating the topology information. In addition, in the TCN message method, an extra time is required depending on the case because the TCN message is once notified to the root bridge and then the root bridge notifies the entire system.
[0017]
For this reason, the STP bridge can detect the disconnection only when the set BPDU that is received periodically does not come. As shown in FIG. 8, the time required from the occurrence of disconnection to the return to topology is the transfer delay time T until the entire system notices the disconnection. _fwd Is the time T required for topology reconstruction _fwd × 2, theoretically the transfer delay time T _fwd × 3.
[0018]
Note that RSTP (Rapid Spanning Tree Protocol) has been proposed in which STP is improved to shorten the topology convergence time. The RSTP states that by adding some conditions under which a port can transit to the forwarding state, the time can be reduced to the forwarding state only in a case where there is no concern about occurrence of a loop. However, even in this RSTP, it is presumed that the transition condition is limited, and when the number of nodes becomes large, such as several thousands, it is not possible to transition to the forwarding state without waiting time.
[0019]
As a method for solving the above problems, the applicant of the present application has proposed an information transfer system and a logical topology construction method of the information transfer system which can eliminate the waiting time for bringing the port into the forwarding state and can speed up the topology construction and the topology restoration. It has already been proposed (see Patent Document 1).
[0020]
In this system, each node Ni (i is a node ID) is configured by a functional block shown in FIG. The main functional blocks are a port state control block (blocking port drift mechanism) 1 and a topology information updater (re-arbitration trigger mechanism) 2. These are described below.
[0021]
(A) Blocking port drift mechanism
Neighboring nodes exchange topology information with each other through transmission and reception of an arbitrated RPDU (Ring Protocol Data Unit), determine the hierarchical relationship between the nodes by comparing their priorities, and finally determine a port role. . In FIG. 9, the topology information buffers 3L and 3R transmit / receive arbitration RPDUs to / from adjacent nodes via the ports L or R at both ends of the node Ni, and the received topology information determines the priority in the topology priority determination algorithm 4, The determined port roles are stored in the port role storage units 5L and 5R.
[0022]
The arbitration RPDU is an RTP-specific frame in which topology information and various flags are stored. When the port transmitting the arbitration RPDU is in a blocking state, the arbitration RPDU is set with a blocking request flag and transmitted (referred to as a blocking request message).
[0023]
If the arbitration caused by the blocking request message is lost, the port state control block 1 accepts the blocking request, shifts the port to the blocking state, and returns an arbitration RPDU with the blocking request acceptance flag set to the upper node. Yes (blocking request acceptance message). Then, upon receiving the blocking request acceptance message, the receiving port immediately transitions to the forwarding state.
[0024]
The repetition of this arbitration operation ensures that the blocking port drifts downstream. Eventually, the loop of the ring topology is broken by the node farthest from the master node continuing to block.
[0025]
The arbitrated port state is set in the setting units 6L and 6R, stored in the switching algorithm & address table 7, and used for frame transfer control between adjacent ports.
[0026]
(B) Re-arbitration trigger mechanism
Each node monitors the link state by hardware, and if a communication failure is detected due to link down or the like, the port is immediately fixed to the blocking state by the re-arbitration trigger mechanism 2 and the arbitration trigger RPDU is transmitted (topology restoration message ). Next, the node clears its own topology information and starts arbitration as the master node.
[0027]
The node that has received the topology restoration message immediately transfers the message to the opposite port, clears its own topology information, and starts arbitration as a master node.
[0028]
The topology restoration message is transferred on the ring, and the arbitration operation is triggered one after another by clearing the topology information of the nodes on the route, and disappears when the disconnection point is reached.
[0029]
By circulating the topology restoration message quickly, it is possible to minimize the delay in starting arbitration.
[0030]
[Non-patent document 1]
ISO / IEC Final DIS 15802-3 (IEEE P802D / D17)-Chapter 8 The spanning tree algorithm and protocol
[0031]
[Patent Document 1]
Japanese Patent Application No. 2002-293210
[0032]
[Problems to be solved by the invention]
In the conventional RTP, an error may occur in the system or the function may be stopped. Specifically, there are the following problems.
[0033]
(1) Node down
In the conventional RTP, the detection of the ring-down is the only trigger of the topology recovery operation. However, when a certain node dies without a ring-down (due to a device abnormality or the like), the adjacent node cannot detect a topology change and cannot release logical blocking. This will be described with reference to FIGS. In the drawing, "●-“ "marks indicate blocking ports. Further, ““ ”(upper port) and“ ● ”(lower port) indicate forwarding ports.
[0034]
FIG. 10A shows a case where normal topology restoration is possible, and the node N ₆ Is turned off [step: 0] and the link to the adjacent node is cut off [step: 1]. ₁ And N ₃ Can detect topological changes. In this case, the logical topology is reconstructed by the operation of the re-arbitration trigger mechanism, and the logical blocking can be released.
[0035]
On the other hand, in FIG. 10B, the node N ₆ Node N without ring-down ₆ Is unable to perform the transfer process [step: 0], the adjacent node is the node N ₆ And the logical topology is not restored forever.
[0036]
(2) Message loss
When the RPDU is destroyed by noise or the like and the partner node cannot receive it, the protocol operation stops there. An example in which the logical topology construction is delayed when the arbitration RPDU is lost will be described below.
[0037]
As shown in FIG. ₂ In a state [step: 0] in which a logical topology in which is the master node is constructed, the node N ₁ Joins the existing topology [step: 1], except that node N ₁ Priority is node N ₂ And higher. At this time, the node N ₁ And node N ₆ Arbitration RPDUs are exchanged between ₁ RPDU transmitted by node N ₆ [Step: 2], the node N ₁ Is the node N ₆ Receives the arbitration RPDU from the other node, but does not follow the RPDU because it wins the arbitration. ₆ ) Acknowledges the defeat in the arbitration and notifies the reception of the logical blocking [step: 3]. At this time, the node N ₆ Is the partner node (N ₁ )) And waits for an arbitration RPDU. In other words, the parties are in a waiting state, and the arbitration will not proceed further unless an event urging the topology construction occurs.
[0038]
(3) Station cache clear
If the blocking port drifts, it may be necessary to change the transfer port of the general frame. To update the forwarding port, the station cache must be cleared. However, in the current method, the frame may not reach the destination because the station cache is not cleared.
[0039]
This example will be described with reference to FIG. In [step: 0], node N ₁ And node N ₃ When a general frame is transferred between, the frame is relayed along the path indicated by the arrow in the figure. Therefore, node N ₁ Switching device is node N ₃ Frame addressed to port P _1L Transfer to Node N ₃ Switching device is node N ₁ Frame addressed to port P _3R Transfer to
[0040]
In this situation, it is assumed that the logical blocking port has drifted and has become [step: 1]. At least node N ₃ Can know the drift of the logical blocking port, and can clear the station cache. ₁ Frame addressed to port P _3R Transfer to "" can be discarded. As a result, node N ₃ To node N ₁ The frame addressed to it is immediately transferred to the new path. On the other hand, node N ₁ Cannot know the drift of the logical blocking port.
[0041]
Therefore, even after the logical blocking port drifts, the node N ₃ Frame addressed to port P _1L Will be transferred to Port P _1L Node N forwarded to ₃ Although the frame addressed to is forwarded, the node N ₃ Port P of _3R Will be blocked.
[0042]
"Node N ₃ Frame addressed to port P _1L It takes an address aging time until the learning of `` transfer to '' is cleared, but since this value is usually set in the order of minutes, regardless of the established communication path, It takes a considerable amount of time before frames can be transferred.
[0043]
(4) Incorrect address learning
The source address of the RPDU is learned, and the general frame is not transferred to the port to which the frame should be transferred.
[0044]
For example, as shown in FIG. ₅ From node _N In the state [step: 0] in which the general frame is transferred to the node N2, ₅ Switching device is "Node N ₂ "To port P _5L To transfer to. "
[0045]
At this time, the node N ₂ To node N ₅ It is assumed that some RPDU frame has been transferred [step: 1]. At this time, the node N ₅ The switching device maps the source address of the frame and the receiving port. As a result, node N ₅ The switching device of "Node N ₂ Frame addressed to port P _5L Discard the previous learning of "transfer to ₂ Frame addressed to port P _5R Transfer to "" and learn again.
[0046]
As a result, node N ₅ The switching devices of node N ₂ Frame addressed to port P _5R [Step: 2]. This allows port P _5R Transmitted from node N ₂ Is blocked by the logical blocking port.
[0047]
(5) Infinite tour of messages
If the nodes are connected in a ring, the message once transmitted is continuously transferred on the ring endlessly. An example in which the re-arbitration trigger RPDU makes an infinite round on the ring will be described below.
[0048]
As shown in FIG. ₁ And N ₆ It is assumed that disconnection has occurred between [step: 0]. Thereby, the node N ₁ And N ₆ Sends a re-arbitration trigger RPDU [step: 1]. This transmission is indicated by "Healing Message".
[0049]
At this time, if the disconnection point returns while the re-arbitration trigger RPDU circulates on the ring topology [step: 2], the re-arbitration trigger RPDU passes through the original disconnection point and continues to circulate on the ring topology. [Step: 3]. In this case, the circulation of the re-arbitration trigger RPDU is continuously continued unless a new disconnection occurs somewhere.
[0050]
(6) Master node change
For example, assume that the ID of the existing master node is changed and reset. If ring-down does not occur due to this reset, the old master node ID remains on the topology, preventing correct arbitration operation.
[0051]
This example will be described with reference to FIG. Node N ₁ Is a master node and a logical topology is constructed [step: 0]. In this state, the node N ₁ Is changed from 1 to 7 [step: 1]. At this time, the node N ₁ The port is blocked once to restart the protocol operation from 1.
[0052]
With this blocking, node N ₁ Is the adjacent node N ₄ , N ₆ And arbitration with [step: 2]. As a result, node N ₁ Is infected with the topology information that he has spread himself before [step: 3]. As a result, the logical mastering node is absent and the logical topology is not correctly constructed, so that the logical blocking port drift mechanism does not operate normally (the logical blocking port does not drift downstream in the topology).
[0053]
As a result, the logical blocking ports are not aggregated at the end of the ring, and the logical topology is divided.
[0054]
(7) Measures against lack of logical blocking port
Although it is impossible as long as the protocol is operating normally, it is assumed that an erroneous blocking response message is generated due to noise or the like, and all ports on the ring are in a forwarding state. If this situation occurs, the blocking drift mechanism becomes completely useless, and unless a ring-down occurs somewhere, the state of zero blocking is continued, which is dangerous.
[0055]
An object of the present invention is to provide an information transfer system that can prevent the occurrence of system abnormalities and the suspension of functions due to the above-described causes.
[0056]
[Means for Solving the Problems]
The present invention is characterized by the following system and logical topology construction method in order to solve the above problems.
[0057]
(System invention)
(1) In order to transfer information between a large number of frame transfer devices, each frame transfer device is used as a node, and unnecessary links are logically disconnected from a topology physically including a loop, and a logical connection between each node without a loop is provided. A port state control mechanism for constructing a topology, and a re-arbitration trigger mechanism for reconstructing a logical topology when a communication failure occurs between the nodes,
The port state control mechanism includes means for issuing a blocking request from a port in a blocking state at each node in the ring topology, and causing a node that has lost arbitration due to this request to transition to the blocking state and receive a blocking request from an upper node. Means, and the node notified of the blocking request reception comprises means for immediately canceling the blocking state and transitioning to the forwarding state,
The re-arbitration trigger mechanism sets a port of a node that has detected a communication failure in a blocking state in hardware, clears topology information, and transfers a topology restoration message to a next node. Means for discarding the topology information and becoming a master node and starting arbitration for the next node, comprising:
The master node exchanges topology information with each other, determines the hierarchical relationship between the nodes by comparing the priorities, and finally provides a means for periodically transmitting an arbitration RPDU for finally determining a port role,
The slave node is characterized in that a means is provided for immediately returning to the master node when the arbitration RPDU from the master node does not arrive for a predetermined time or longer, regardless of whether or not a re-arbitration trigger RPDU has been received.
[0058]
(2) Among the nodes, the sender of the message is provided with a means for periodically retransmitting the message until receiving an incoming response to the transmitted message;
The message receiver is provided with means for returning an incoming response to the received message,
An ACK flag is provided in the arbitration RPDU, and means for setting this flag when a message is returned as an incoming response is provided.
[0059]
(3) When the node changes the port state, a means is provided for transmitting a cache clear RPDU urging another node to clear the cache under the following conditions:
If the node whose port state has been changed is the master node, a cache clear RPDU is transmitted from the ring port in the forwarding state.
[0060]
If the node whose port state has been changed is a slave node, a cache clear RPDU is transmitted from the upper port.
[0061]
The node receiving the cache clear RPDU,
Means for transferring the cache clear RPDU when the receiving port is in the forwarding state and the transfer destination port is in the forwarding state, when the node receiving the cache clear RPDU is the master node,
When the node receiving the cache clear RPDU is a slave node, means for transferring the cache clear RPDU when the receiving port is in the forwarding state and the transfer destination port is the upper port is provided.
[0062]
(4) The ring terminal port farthest from the master node in the ring topology is:
-Provide an "end flag" for notifying the arbitration RPDU that it is a ring end port;
The opposite node that has received the arbitrated RPDU is provided with a means for inhibiting the learning of the address of the receiving port.
[0063]
(5) The port that has received the arbitrated RPDU is provided with means for transferring the arbitrated RPDU when the transfer destination port is not the upper port and the transfer destination port is not the terminal port;
The re-arbitration trigger RPDU or the cache clear RPDU is provided with a unit for transmitting by attaching a parameter of "source node ID" and a parameter of "lifetime",
A node that first transmits the re-arbitration trigger RPDU or the cache clear RPDU includes a unit that writes its own unique node ID in the “source node ID” and a maximum number of ring nodes in the “lifetime”;
The node that has received the re-arbitration trigger RPDU or the cache clear RPDU determines that the “source node ID” does not match its own unique node ID and that the “lifetime” is decremented by 1 or more. A means for transferring an arbitration trigger RPDU or a cache clear RPDU is provided.
[0064]
(6) The slave node constantly monitors a change in the priority of the topology information notified by the arbitration RPDU received from the upper port, and receives topology information having a lower priority than the topology information received last time. In that case, depending on the pattern,
If the value of the master node ID does not change and the distance value increases, the node returns to the master node and sends a re-arbitration trigger RPDU.
[0065]
-If the value of the master node ID increases but the priority is higher than its own unique node ID, it depends on the topology information.
[0066]
When the value of the master node ID increases and the priority becomes lower than the own unique node ID, the process returns to the master node.
Is provided.
[0067]
(7) The node at the end of the ring is provided with a means for spontaneously blocking one of the ring ports when it recognizes that it is the end of the ring, even if it has not received a blocking request from an adjacent node. It is characterized by the following.
[0068]
(Invention of method)
(8) In order to transfer information between a large number of frame transfer devices, each frame transfer device is used as a node, and unnecessary links are logically disconnected from a topology including a physical loop, and a logical connection between each node without a loop is provided. A port state control mechanism for constructing a topology, and a re-arbitration trigger mechanism for reconstructing a logical topology when a communication failure occurs between the nodes,
The port state control mechanism includes means for issuing a blocking request from a port in a blocking state at each node in the ring topology, and causing a node that has lost arbitration due to this request to transition to the blocking state and receive a blocking request from an upper node. Means, and the node notified of the blocking request reception comprises means for immediately canceling the blocking state and transitioning to the forwarding state,
The re-arbitration trigger mechanism sets a port of a node that has detected a communication failure in a blocking state in hardware, clears topology information, and transfers a topology restoration message to a next node. Means for constructing a logical topology of an information transfer system, comprising means for discarding topology information and becoming a master node and starting arbitration for the next node,
The master node exchanges topology information with each other, determines a hierarchical relationship between the nodes by comparing their priorities, and finally provides a process of periodically transmitting an arbitration RPDU for determining a port role,
The slave node is characterized in that, when the arbitration RPDU from the master node does not arrive for a predetermined time or more, a step of immediately returning to the master node is provided regardless of whether or not a re-arbitration trigger RPDU is received.
[0069]
(9) Among the nodes, the message sender has a process of periodically retransmitting the message until receiving an incoming response to the transmitted message,
The message receiver has a process of returning an incoming response to the received message,
An ACK flag is provided in the arbitration RPDU, and a step of setting this flag when a message is returned as an incoming response is provided.
[0070]
(10) When the node changes the port state, a step of transmitting a cache clear RPDU urging another node to clear the cache under the following conditions is provided:
If the node whose port state has been changed is the master node, a cache clear RPDU is transmitted from the ring port in the forwarding state.
[0071]
If the node whose port state has been changed is a slave node, a cache clear RPDU is transmitted from the upper port.
[0072]
The node receiving the cache clear RPDU,
When the node that has received the cache clear RPDU is the master node, a step of transferring the cache clear RPDU when the receiving port is in the forwarding state and the transfer destination port is in the forwarding state is provided;
When the node that has received the cache clear RPDU is a slave node, a step of transferring the cache clear RPDU when the receiving port is in the forwarding state and the transfer destination port is the upper port is provided.
[0073]
(11) In the ring topology, the ring terminal port farthest from the master node is:
-Provide an "end flag" for notifying the arbitration RPDU that it is a ring end port;
The opposite node that has received the arbitrated RPDU is characterized in that a process for inhibiting address learning of the receiving port is provided.
[0074]
(12) The port receiving the arbitrated RPDU is provided with a step of transferring the arbitrated RPDU when the transfer destination port is not the upper port and the transfer destination port is not the terminal port;
The re-arbitration trigger RPDU or the cache clear RPDU is provided with a process of attaching a parameter of “source node ID” and a parameter of “lifetime” and transmitting the same.
The node that first transmits the re-arbitration trigger RPDU or the cache clear RPDU includes a process of writing its own unique node ID in the “source node ID” and writing a maximum number of ring nodes in the “lifetime”.
The node that has received the re-arbitration trigger RPDU or the cache clear RPDU determines that the “source node ID” does not match its own unique node ID and that the “lifetime” is decremented by 1 or more. A process for transferring an arbitration trigger RPDU or a cache clear RPDU is provided.
[0075]
(13) The slave node constantly monitors the change in the priority of the topology information notified by the arbitration RPDU received from the upper port, and receives the topology information having a lower priority than the previously received topology information. In that case, the following response process depending on the pattern,
If the value of the master node ID does not change and the distance value increases, the node returns to the master node and sends a re-arbitration trigger RPDU.
[0076]
-If the value of the master node ID increases but the priority is higher than its own unique node ID, it depends on the topology information.
[0077]
When the value of the master node ID increases and the priority becomes lower than the own unique node ID, the process returns to the master node.
Is provided.
[0078]
(14) The node at the end of the ring, when recognizing that it is the end of the ring, has a process of spontaneously blocking one of the ring ports even if it has not received a blocking request from an adjacent node. It is characterized by the following.
[0079]
BEST MODE FOR CARRYING OUT THE INVENTION
(Embodiment 1) Node down countermeasures
In this embodiment, the master node periodically sends an arbitration RPDU to indicate that the node is alive. Each slave node constantly monitors the upper port (the ring port closer to the master node among the two ring ports), and discards the current topology information if it does not receive an arbitrated RPDU for a certain period of time, Return to the master node and start arbitration by itself.
[0080]
By adding this function, if the master node goes down, all slave nodes will notice after a certain period of time, and if the slave node goes down, all slaves downstream of that slave will notice after a certain period And the logical topology construction work can be resumed.
[0081]
This example is shown in FIG. 1, and the logical topology for the node down is reconstructed by the following step: 0 to step: 3.
[0082]
In the state of [step: 0], the node N ₁ The logical topology is constructed using as the master node (in the figure, the logical blocking port is P _B Notation). In this state, the slave node N ₆ Is down. However, the adjacent node is node N ₆ Did not notice the unusual.
[0083]
In [step: 1], the slave node N ₄ , N ₅ Continue to receive the arbitrated RPDU from the upper port (indicated by a circle in the figure). On the other hand, slave node N ₃ , N ₂ Is the node N ₆ Does not receive the arbitration RPDU from the upper port because the arbitration RPDU is no longer relayed. Note that node N ₂ The upper port is Node N ₃ Side port.
[0084]
In [step: 2], the slave node N ₃ , N ₂ Since a certain period of time has passed since the arbitration RPDU is no longer received from the upper port, it returns to the master node and starts arbitration.
[0085]
In [step: 3], the arbitration causes the node N ₂ , N ₃ Is the node N ₅ Depends on the topology information via This arbitration also causes the logical blocking port P _B Is the node N ₃ Is drifted down to the lower port.
[0086]
Through the above steps, the logical topology is reconstructed. Therefore, according to the present embodiment, when a certain node stops the RTP protocol operation without causing a ring-down due to a device failure or S / W runaway (that is, the re-arbitration trigger mechanism 2 in the conventional RTP operates normally). In this case, the logical topology is started to be reconstructed after a certain period of time has elapsed, and if there is a bypass route, a communication route that bypasses the failure location is established.
[0087]
(Embodiment 2) Message loss countermeasures
In the present embodiment, as shown in FIG. 2, the transmitting node continues retransmission until a response (ACK) to the message arrives. The receiving node returns an ACK to the message. As a result, even if the message is lost, retransmission is repeated, so that the protocol operation does not stop.
[0088]
This example is shown in FIG. 3, and the following step: 0 to step: 3 prevents message loss.
[0089]
In [step: 0], the node N ₁ And N ₆ Sent arbitration RPDUs respectively, but node N ₁ The one sent from lost.
[0090]
In [step: 1], the node N ₁ Has received the arbitration RPDU, but has won the arbitration, and thus waits for the notification of the subordination from the partner node. Node N ₆ Is waiting for an arbitration RPDU from the partner node.
[0091]
In [step: 2], the node N ₁ , N ₆ Waited for a while, but since there was no reply from the other party, node N ₁ , N ₆ Retransmitted the arbitration RPDU. This time it arrived at the other node without any loss.
[0092]
In [step: 3], the node N ₆ Is the node N ₁ And updates the topology information to the appropriate values and takes over the logical blocking port. Next, node N ₆ Is the node N ₃ Start a new arbitration to relay them to the side. Node N ₁ ACK is returned to inform that the message has been received. Node N with ACK received ₁ Terminates the retransmission of the arbitration RPDU.
[0093]
After that, proceed with arbitration in the same manner. Complete the logical topology. Therefore, according to the present embodiment, even if an arbitration RPDU or a notification RPDU (rearbitration trigger RPDU, cache clear RPDU) is lost during transmission due to the influence of transmission noise or transient overload, message retransmission is performed. Is performed, and the inconvenience of impairing the protocol operation itself is eliminated, although the logical topology convergence time is affected.
[0094]
(Embodiment 3) Station cache clear mechanism
As described in the conventional problem, when the logical blocking port drifts, the transfer route of the general frame may need to be changed. In order to change the route, the address transfer information learned by each node must be once discarded and the address learning must be performed again.
[0095]
Therefore, in the present embodiment, when a certain node changes the port state of its own ring port (that is, when the logical blocking port drifts), the following means A and B for notifying the change to the entire system are newly provided. .
[0096]
(Means A) Issuance of request for address learning redo
The node that has changed the port state transmits a special message (hereinafter, referred to as “cache clear RPDU”) requesting clearing of the station cache according to the following conditions.
[0097]
-When the node is a master node ... A cache clear RPDU is transmitted from the ring port in the forwarding state at that time.
[0098]
When the node is a slave node: The cache clear RPDU is transmitted from the upper port.
[0099]
(Means B) Relay of cache clear RPDU
The node that has received the cache clear RPDU transfers the same message to the opposite ring port according to the following conditions.
[0100]
When the node is a master node: If the received port is in the forwarding state and the transfer destination ring port is in the forwarding state, the cache clear RPDU is transferred.
[0101]
When the node is a slave node: If the received port is in the forwarding state and the transfer destination ring port is the upper port, the cache clear RPDU is transferred.
[0102]
By the above procedure, all nodes connected by the forwarding port at that time are notified of the station cache update.
[0103]
Although the cache clear RPDU is not relayed to the nodes behind the blocking port, the blocking port drifts due to arbitration, and the final blocking port drift is completed, and all nodes on the topology are connected by the forwarding port. All nodes are notified of station cache clear. In other words, not only can the necessary and sufficient station cache be cleared without relaying the cache clear RPDU beyond the blocking port, but unnecessary cache clear RPDU transfer is suppressed.
[0104]
FIG. 4 shows a case where a node N is indicated by a cache clear RPDU (denoted as “CC” in FIG. ₁ Shows a process in which an old transfer route learned previously is cleared.
[0105]
In [step: 0], the blocking drift request is ₆ (The blocking flag of the arbitration RPDU is set and transmitted).
[0106]
In [step: 1], the node N ₃ Accepted the blocking request because it lost the arbitration _3R Was blocked. Node N ₃ Changed the port state (port P _3R Port state: Forwarding → Blocking), the cache clear RPDU is transmitted from the upper port.
[0107]
In [step: 2], the node N ₆ Has received the cache clear RPDU, but ignores it because the receiving port is in a blocking state. Node N ₃ Indicates that the logical blocking has been accepted ₆ (ACK in the figure).
[0108]
In [step: 3], the node N ₆ Received the blocking ACK and released the blocking of the lower port. Node N ₆ Sends a cache clear RPDU from the upper port. Node N ₁ Receives this cache clear RPDU, clears its own station cache, and discards old learning.
[0109]
Thereafter, the cache clear RPDU is sequentially relayed, the station cache of each node on the route is cleared, and the node N ₃ Port P of _3R (The cache clear RPDU received at the blocking port is not transferred).
[0110]
Therefore, according to the present embodiment, when the transfer route of the general frame is changed due to the drift of the logical blocking port, the old address information held by the switching device of each node is quickly cleared, so that the normal The communication state can be restored without waiting for the aging time (generally, the time on the order of minutes) to elapse.
[0111]
(Embodiment 4) Prohibition of address learning
In the present embodiment, the following means is provided to prohibit address learning of a port facing a terminal port (a port that constantly maintains logical blocking in the ring topology) in the ring topology.
[0112]
(Means A) Notification of end port: An "end flag" is provided in the arbitration RPDU, and when the arbitration RPDU is transmitted from the end port, the flag is set.
[0113]
(Means B) Opposite end port ... A node that has received an arbitrated RPDU in which a port end flag is set prohibits address learning of the receiving port.
[0114]
(Means C) Address learning permission: The node that has received the arbitrated RPDU with the terminal flag cleared clears the address learning of the receiving port.
[0115]
FIG. 5 shows an example of the address learning prohibition process.
[0116]
In [step: 0], the node N ₅ To node N ₂ Is a state where a general frame is being transferred. At this time, the node N ₅ Switching device is "Node N ₂ Frame addressed to port P _5L To transfer to. "
[0117]
In [step: 1], the node N ₂ To node N ₅ Assume that some arbitration RPDU frame has been transferred. However, port P _2L Is a terminal port, the terminal flag of the arbitration RPDU is set (in the figure, flag Terminal = 1 is indicated).
[0118]
At this time, the node N ₅ Is the port P because the terminal flag of the received arbitration RPDU is set. _5R Prohibit address learning. Node N ₅ Switching device is port P _5R Has stopped learning the address of node N ₂ Frame addressed to port P _5L The previous learning of "transfer to" remains without being abandoned.
[0119]
In [step: 2], the node N ₅ The switching device of node N ₂ Frame addressed to port P _5L Continue to transfer to. Port P _5L Transmitted from the node N through the conventional route ₂ And is relayed to the final destination.
[0120]
Therefore, according to the present embodiment, the phenomenon that the transfer of the general frame is obstructed can be solved by preventing the erroneous address learning caused by the RPDU.
[0121]
(Embodiment 5) Infinite loop prevention of message
In the present embodiment, when a ring topology is configured, RPDU is prevented from going infinitely on the ring as follows.
[0122]
(A) Arbitration RPDU
The transfer conditions when an arbitrated RPDU is received are set such that the transfer destination port is not an upper port and that the transfer destination port is not a terminal port.
[0123]
According to this condition, as shown in FIG. 6A, the arbitrated RPDU is always transferred in the upper order → lower direction, and the transfer stops at the node located farthest from the master node on the ring topology (lower order → upper direction). Will not be forwarded to
[0124]
(B) Re-arbitration trigger RPDU and cache clear RPDU
The following two parameters are attached to the re-arbitration trigger RPDU and the cache clear RPDU (these RPDUs are hereinafter referred to as "notification RPDUs").
[0125]
-Source node ID
·lifespan
First, the node transmitting the notification RPDU writes its own unique node ID in the transmission source node ID and the maximum number of ring nodes in the lifetime. The node that has received the notification RPDU transfers the RPDU only when the following conditions are satisfied.
[0126]
(Condition A) The source node ID is compared with its own unique node ID, and they do not match. (Notification sent by yourself)
(Condition B) The life is decremented (decrease the value by 1) and the value is 1 or more.
[0127]
The condition A imposes the tail of the notification RPDU on the transmission source node ID. Due to this restriction, the notification RPDU disappears when it makes one round of the ring, as shown in FIG. 6B. .
[0128]
Condition b is that when the unique node ID of the node N is changed before a certain node N transmits the notification RPDU and the RPDU makes one round on the ring, that is, there is a node that cleans up the notification RPDU. This is a restriction to prevent an infinite tour even if it disappears.
[0129]
Therefore, according to the present embodiment, it is possible to prevent the RPDU from infinitely circulating in the ring topology, and to suppress the generation of unnecessary traffic.
[0130]
(Embodiment 6) Master node ID change countermeasures
In the present embodiment, the following rule is set to detect an invalid master node ID as soon as possible and prevent it from being spread. This rule is applied to the slave node.
[0131]
Legend: a certain slave node N _S Representative topology information (master node ID, distance, source node ID) _D (T _D = (M _D , D _D , S _D ))). Here, node N _S Has newly received an arbitrated RPDU at an upper port. Topology information T stored in this arbitration RPDU _R Is hereinafter referred to as reception topology information (T _R = (M _R , D _R , S _R )). Node N _S The unique node ID of U _S And
[0132]
(Condition 1) When the received topology information is superior to or equal to the representative topology information: The representative topology information is recalculated based on the received topology information (process as in the related art).
[0133]
(Condition 2) When the received topology information is inferior to the representative topology information,
When the value of the representative master node ID is equal to the value of the receiving master node ID under the condition 2 (M _D = M _R ), The node Ns returns to the master node and transmits a re-arbitration trigger RPDU.
[0134]
In the condition 2, the values of the representative master node ID and the receiving master node ID are different (M _D ≠ M _R ) And the value of the receiving master node ID is smaller than the value of the unique node ID of the node Ns (M _R <U _S ) ... Recalculate the representative topology information based on the received topology information (process as in the past).
[0135]
In the condition 2, the value of the reception master node ID is the node N _S Is larger than the value of the unique node ID (M _R > U _S ) ... Node N _S Returns to the master node and does not send a re-arbitration trigger RPDU.
[0136]
Therefore, according to the present embodiment, when the unique node ID of the node currently serving as the master node is changed and its value becomes larger than before (that is, the priority becomes lower than before), The topology information no longer remains in the topology, and the logical topology is correctly reconstructed.
[0137]
(Embodiment 7) Spontaneous blocking
In the present embodiment, the node located at the end of the ring spontaneously blocks one of the ring ports (even if there is no blocking request from the upper node). Also, even if a blocking response is received at a port where the ring terminal node is blocking, the blocking is not released.
[0138]
Therefore, according to the present embodiment, even if all the blocking ports on the link topology have been released due to some cause (for example, mixing of a forged RPDU), the node located at the end of the ring is automatically activated. Maintaining the logical blocking can prevent the loop state from being left.
[0139]
【The invention's effect】
As described above, according to the present invention, in an information transfer system using a conventional ring topology protocol, improved resistance to node down, improved resistance to protocol message loss, secure station cache clear, prevention of erroneous address learning, and It is possible to prevent infinite patrol, secure correct arbitration for master node change, and secure voluntary blocking.
[Brief description of the drawings]
FIG. 1 is a process diagram of a node down countermeasure according to a first embodiment of the present invention.
FIG. 2 is a diagram illustrating a relationship between a message and an ACK according to the second embodiment of the present invention.
FIG. 3 is a processing diagram of a message loss countermeasure according to the second embodiment of the present invention.
FIG. 4 is a diagram illustrating a cache clearing process using a cache clear RPDU according to a third embodiment of the present invention.
FIG. 5 is a diagram illustrating an address learning prohibition process according to a fourth embodiment of the present invention.
FIG. 6 is an explanatory diagram showing prevention of infinite message circulation according to the fifth embodiment of the present invention.
FIG. 7 is a transition diagram of the port state of the STP bridge.
FIG. 8 is a process diagram of STP disconnection recovery.
FIG. 9 is a functional block diagram of an RTP.
FIG. 10 is an example of node down in RTP.
FIG. 11 is an example of stagnation of arbitration due to message loss in RTP.
FIG. 12 is an example of a change of a transfer port due to a drift of a logical blocking port in RTP.
FIG. 13 shows a mechanism of occurrence of a transfer failure due to incorrect address learning in RTP.
FIG. 14 is an example of infinite circulation of a message in RTP.
FIG. 15 is an example of a residual phenomenon of old topology information in RTP.
[Explanation of symbols]
1: Port state control block
2. Topology information updater
3L, 3R: Topology information buffer
4. Topology superiority / disadvantage determination algorithm
5L, 5R: Port roll storage unit
6L, 6R: Port state storage unit
7: Switching algorithm & address table

Claims (14)

In order to transfer information between a large number of frame transfer devices, each frame transfer device is used as a node, and unnecessary links are logically disconnected from the topology that physically includes loops, creating a logical topology without loops between each node. A port state control mechanism, and a re-arbitration trigger mechanism for reconstructing a logical topology when a communication failure occurs between the nodes,
The port state control mechanism includes means for issuing a blocking request from a port in a blocking state at each node in the ring topology, and causing a node that has lost arbitration due to this request to transition to the blocking state and receive a blocking request from an upper node. Means, and the node notified of the blocking request reception comprises means for immediately canceling the blocking state and transitioning to the forwarding state,
The re-arbitration trigger mechanism sets a port of a node that has detected a communication failure in a blocking state in hardware, clears topology information, and transfers a topology restoration message to a next node. Means for discarding the topology information and becoming a master node and starting arbitration for the next node, comprising:
The master node exchanges topology information with each other, determines the hierarchical relationship between the nodes by comparing the priorities, and finally provides a means for periodically transmitting an arbitration RPDU for finally determining a port role,
The information transfer system according to claim 1, wherein said slave node is provided with a means for immediately returning to the master node when the arbitration RPDU from the master node does not arrive for a predetermined time or longer, regardless of whether a re-arbitration trigger RPDU is received. . Among the nodes, a message sender is provided with a means for periodically resending a message until an incoming response to the transmitted message is received,
The message receiver is provided with means for returning an incoming response to the received message,
2. The information transfer system according to claim 1, wherein an "ACK flag" is provided in the arbitration RPDU, and means for setting this flag when a message is returned as an incoming response is provided. When the node changes the port state, a means for transmitting a cache clear RPDU urging another node to clear the cache under the following conditions is provided:
If the node whose port state has been changed is the master node, a cache clear RPDU is transmitted from the ring port in the forwarding state.
If the node whose port state has been changed is a slave node, a cache clear RPDU is transmitted from the upper port.
The node receiving the cache clear RPDU,
Means for transferring the cache clear RPDU when the receiving port is in the forwarding state and the transfer destination port is in the forwarding state, when the node receiving the cache clear RPDU is the master node,
When the node receiving the cache clear RPDU is a slave node, means for transferring the cache clear RPDU when the receiving port is in the forwarding state and the transfer destination port is the upper port is provided. Item 3. The information transfer system according to item 1 or 2. The ring terminal port farthest from the master node in the ring topology is
-Provide an "end flag" for notifying the arbitration RPDU that it is a ring end port;
The information transfer system according to any one of claims 1 to 3, wherein the opposite node that has received the arbitrated RPDU is provided with a unit that prohibits address learning of a receiving port. The port receiving the arbitration RPDU is provided with means for transferring the arbitration RPDU when the transfer destination port is not the upper port and the transfer destination port is not the terminal port,
The re-arbitration trigger RPDU or the cache clear RPDU is provided with a unit for transmitting by attaching a parameter of "source node ID" and a parameter of "lifetime",
A node that first transmits the re-arbitration trigger RPDU or the cache clear RPDU includes a unit that writes its own unique node ID in the “source node ID” and a maximum number of ring nodes in the “lifetime”;
The node that has received the re-arbitration trigger RPDU or the cache clear RPDU determines that the “source node ID” does not match its own unique node ID and that the “lifetime” is decremented by 1 or more. 5. The information transfer system according to claim 1, further comprising means for transferring an arbitration trigger RPDU or a cache clear RPDU. The slave node continuously monitors the change in the priority of the topology information notified by the arbitration RPDU received from the upper port, and when receiving the topology information having a lower priority than the previously received topology information, Depending on the pattern,
If the value of the master node ID does not change and the distance value increases, the node returns to the master node and sends a re-arbitration trigger RPDU.
-If the value of the master node ID increases but the priority is higher than its own unique node ID, it depends on the topology information.
When the value of the master node ID increases and the priority becomes lower than the own unique node ID, the process returns to the master node.
The information transfer system according to any one of claims 1 to 5, further comprising: The node at the end of the ring, when recognizing that the node is the end of the ring, has a means for spontaneously blocking one of the ring ports even if it has not received a blocking request from an adjacent node. The information transfer system according to any one of claims 1 to 6, wherein In order to transfer information between a large number of frame transfer devices, each frame transfer device is used as a node, and unnecessary links are logically disconnected from the topology that physically includes loops, creating a logical topology without loops between each node. A port state control mechanism, and a re-arbitration trigger mechanism for reconstructing a logical topology when a communication failure occurs between the nodes,
The port state control mechanism includes means for issuing a blocking request from a port in a blocking state at each node in the ring topology, and causing a node that has lost arbitration due to this request to transition to the blocking state and receive a blocking request from an upper node. Means, and the node notified of the blocking request reception comprises means for immediately canceling the blocking state and transitioning to the forwarding state,
The re-arbitration trigger mechanism sets a port of a node that has detected a communication failure in a blocking state in hardware, clears topology information, and transfers a topology restoration message to a next node. Means for constructing a logical topology of an information transfer system, comprising means for discarding topology information and becoming a master node and starting arbitration for the next node,
The master node exchanges topology information with each other, determines a hierarchical relationship between the nodes by comparing their priorities, and finally provides a process of periodically transmitting an arbitration RPDU for determining a port role,
An information transfer system, wherein the slave node immediately returns to the master node when the arbitration RPDU from the master node does not arrive for a predetermined time or longer, regardless of whether or not a re-arbitration trigger RPDU is received. How to build a logical topology. Among the nodes, a message sender is provided with a step of periodically retransmitting a message until receiving an incoming response to the transmitted message,
The message receiver has a process of returning an incoming response to the received message,
9. The method according to claim 8, further comprising the step of providing an "ACK flag" in the arbitration RPDU and setting the flag when a message is returned as an incoming response. . When the node changes the port state, the node provides a process of transmitting a cache clear RPDU urging another node to clear the cache under the following conditions,
If the node whose port state has been changed is the master node, a cache clear RPDU is transmitted from the ring port in the forwarding state.
If the node whose port state has been changed is a slave node, a cache clear RPDU is transmitted from the upper port.
The node receiving the cache clear RPDU,
When the node that has received the cache clear RPDU is the master node, a step of transferring the cache clear RPDU when the receiving port is in the forwarding state and the transfer destination port is in the forwarding state is provided;
When the node receiving the cache clear RPDU is a slave node, a step of transferring the cache clear RPDU when the receiving port is in the forwarding state and the transfer destination port is the upper port is provided. Item 10. The method for constructing a logical topology of an information transfer system according to item 8 or 9. The ring terminal port farthest from the master node in the ring topology is
-Provide an "end flag" for notifying the arbitration RPDU that it is a ring end port;
The method according to any one of claims 8 to 10, further comprising a step of prohibiting a counterpart node that has received the arbitrated RPDU from learning the address of the receiving port. The port receiving the arbitration RPDU includes a step of transferring the arbitration RPDU when the transfer destination port is not the upper port and the transfer destination port is not the terminal port,
The re-arbitration trigger RPDU or the cache clear RPDU is provided with a process of attaching a parameter of “source node ID” and a parameter of “lifetime” and transmitting the same.
The node that first transmits the re-arbitration trigger RPDU or the cache clear RPDU includes a process of writing its own unique node ID in the “source node ID” and writing a maximum number of ring nodes in the “lifetime”.
The node that has received the re-arbitration trigger RPDU or the cache clear RPDU determines that the “source node ID” does not match its own unique node ID and that the “lifetime” is decremented by 1 or more. 12. The method according to claim 8, further comprising the step of transferring an arbitration trigger RPDU or a cache clear RPDU. The slave node continuously monitors the change in the priority of the topology information notified by the arbitration RPDU received from the upper port, and when receiving the topology information having a lower priority than the previously received topology information, Depending on the pattern, the following response process,
If the value of the master node ID does not change and the distance value increases, the node returns to the master node and sends a re-arbitration trigger RPDU.
-If the value of the master node ID increases but the priority is higher than its own unique node ID, it depends on the topology information.
When the value of the master node ID increases and the priority becomes lower than the own unique node ID, the process returns to the master node.
The logical topology construction method for an information transfer system according to any one of claims 8 to 12, further comprising: The node at the end of the ring, when recognizing that it is the end of the ring, has a step of spontaneously blocking one of the ring ports even if it has not received a blocking request from an adjacent node. The method for constructing a logical topology of an information transfer system according to any one of claims 8 to 13.

Images