JP4351505B2 - Packet transfer method, packet transfer system, and packet transfer apparatus - Google Patents

Description

  The present invention relates to a packet transfer method, a packet transfer system, and a packet transfer apparatus, and more particularly, when a source address of a received packet is learned in association with its reception path, and a packet is received with the source address as a destination address. Relates to a packet transfer method, a packet transfer system, and a packet transfer apparatus to which a system for determining a transfer route to be applied based on the learning result is applied.

  As a packet transfer method via the network, the MAC address of the source of the received IP packet is learned along with its reception route, and the corresponding learning result is used when an IP packet having the same MAC address as the destination address is received Then, there is a method for determining the transmission route. In this case, the transmission path cannot be specified when transferring an IP packet having an unlearned MAC address as a destination address in a node that relays each IP packet. Therefore, in such a case, broadcast transmission (that is, “flooding”) is performed for all nodes. At that time, path learning is performed by storing the MAC address (SA) of the transmission source of the IP packet and the port that received the IP packet in a buffer of each node.

  1 and 2 show this state. In the example described below shown in FIG. 1 and the like, it is assumed that each of the nodes 50-0, 50-1, and 50-2 has a single communication card # 0, # 1, and # 2. ing. When the node 50-0 receives an IP packet from the terminal A10 at the port # 0 of the card # 0 (steps S1 and P1), the source MAC address (SA) in the MAC table of the card Learning is performed by storing “A” in a predetermined buffer in association with information (route information) of the receiving card # 0 and port # 0 at that time (step S2). In this case, when a learning result for the same MAC address already exists in the MAC table, the information is overwritten.

  Next, it is searched whether there is a learning result in its own MAC table for the destination address (DA, in this case, MAC address “B”) of the received IP packet (step S3). If the search result is a miss-hit (No), the IP packet is broadcast (flooded) to all nodes (in this example, the nodes 50-1 and 50-2) (steps S5 and P2). ). On the other hand, as a result of the search in step S3, if the learning result regarding the corresponding destination MAC address exists in the buffer (Yes), the IP packet is transmitted by the transfer path included in the learning result, that is, the card number and the port number ( Step S4). In this case, there is no need for flooding.

  The same learning operation is also performed when the IP packet issued from the node 50-0 reaches the card # 2 of the node 50-2 that passes through to the corresponding destination terminal B20. As a result, as shown in FIG. 1, the buffer (MAC table) of the card # 2 is associated with the MAC address “A” of the source terminal 10 and the source route information at that time, that is, the node 50-0. Information of card # 0 and port # 0 is stored as a learning result.

  As described above, once an IP packet is received, the source MAC address (SA) and reception path information (card number, port number) are stored in association with each other (that is, learned), so that the next time When an IP packet having the MAC address as the destination (DA) is received, the transfer path can be determined by using the learning result. As a result, as long as the learning result regarding the destination MAC address (DA) of the received packet exists in the buffer, there is no need for flooding because it is possible to uniquely determine the transfer route to be applied to the packet using that information, Thus, it is possible to minimize the traffic of the connected LAN.

  That is, in FIG. 3, when the IP packet from the terminal B20 is received at the card # 2 and the port # 1 of the node 50-2 (steps S1 and P4), as described above, from the terminal A10 (see FIG. 1) already. As the learning result when relaying the IP packet (steps P2 and P3), the reception path information (card # 0, port # 0) is stored in association with the MAC address of the terminal A. A hit is obtained for “A” of the destination address (DA) of the IP packet (Yes in step S3). In this case, the corresponding route, that is, the card # 0 and the port # 0 are applied to the IP packet based on the hit result, and is transferred from the node 50-2 to the card # 0 of the node 50-0 (step P5). ). In this case, the card # 0 of the node 50-0 has a learning result regarding the MAC address “A” which is the corresponding destination address at the time of relaying in steps P1 and P2 in FIG. Can be uniquely identified, and the IP packet can be transferred to the terminal A10 without flooding (step P6).

  Note that the series of packet transfer operations described with reference to FIGS. 1 and 3 are further illustrated in FIG. 4 as time elapses. Step numbers P1 to P6 shown in FIG. 4 correspond to P1 to P6 in FIGS. 1 and 3, respectively.

  Also, in such a system, from the viewpoint of effective use of the limited buffer storage capacity, if there is no DA search hit for the same address for a certain period of time after learning, the learning result regarding the MAC address is deleted. . The operation leading to this deletion is referred to as “aging”. Specifically, as shown in FIG. 5, the card buffer of each node has a “hit bit” corresponding to the learned MAC address, and the hit bit = “1” at the time of learning. In the first aging process, the hit bit is cleared (that is, “0” is set, and steps S11 and S12). In this case, the learning result is not yet deleted. If the hit bit remains “0” during the next aging process (No in steps S14, S15, and S11), the corresponding MAC address is deleted from the buffer (step S13). On the other hand, when overwriting is performed before that, or when the hit search is performed in the DA search (that is, the destination search operation in step S3 in FIG. 2), the hit bit is set to “1” again. As a result, Yes in step S11, S12). With such a configuration, an address currently being used (that is, communication is being executed) is prevented from being deleted from the buffer.

  This aging process will be further described with reference to FIGS. Steps P1 to P6 in FIG. 7 are the same process as steps P1 to P6 in FIG. Each time an IP packet is transferred, the MAC address “A” is learned at the card # 0 in steps P1 and P2, and the hit bit is set to “1”. Similarly, at steps P2 and P3, the card # 1 and card # 2 learn about the MAC address “A”. Similarly, in steps P4, P5, P5 and P6, the MAC address “B” is learned in the card # 2 and the card # 0.

  Thereafter, as shown in the uppermost row of FIG. 8, the hit bits are cleared for the MAC addresses “A” and “B” in the cards # 0, # 1, and # 2 by the first aging process (step S12). . Assume that the next IP packet addressed to the terminal B20 is issued from the terminal A10 before the second aging process (step P7). Upon receiving this, the card # 0 learns by overwriting the transmission source MAC address “A”, and at the same time, the destination MAC address “B” becomes the search target in the destination search. Therefore, the hit bit is returned to “1” for each of the MAC addresses “A” and “B” once cleared to “0” as described above.

  Then, the IP packet is transferred to the card # 2 according to the learning result for the MAC address “B” hit in the destination search (step P8). Also in the card # 2, the hit bit is returned to “1” for each of the MAC addresses “A” and “B” once cleared to “0” by the same operation as described above. Then, it is assumed that the next aging processing timing has arrived without receiving an IP packet related to any of the terminals A, B10 and 20. In this case, as shown at the bottom of FIG. 8, in each of the cards # 0 and # 2, the hit bits for the MAC addresses “A” and “B” are set to “ Since they are returned to “1”, they are cleared to “0” again, but the corresponding learning results are not yet erased at this point (step S12). On the other hand, for the card # 1, the corresponding IP packet does not pass during that time, and the hit bits for the MAC addresses “A” and “B” remain “0”. (Step S13).

  Next, the link aggregation function will be described. This link aggregation function is a function that uses a plurality of ports as a single virtual high-speed port. With this function, the effect of improving the bandwidth can be obtained substantially. The virtual port in this case is called a trunk, and when this trunk is used (called trunking), when performing flooding when an address has not been learned as described above, the trunk of a large number of ports of the trunk is determined by a predetermined calculation. From this, one representative port is selected, and actual packet transmission is performed at this representative port.

  When a node that receives an IP packet via a trunk performs route learning for the source address of the packet, instead of learning the port number at the time of reception, information on the trunk that includes it is stored as a learning result. To do. Thereafter, when receiving an IP packet having a destination passing through the trunk, one route is determined from among the routes constituting the trunk by a predetermined calculation method independent for each hardware of the communication card, and Apply this decision path. By adopting such a configuration, it is possible to prevent different routes in the same trunk from being applied to the same packet.

  The state of application of the link aggregation function will be described with reference to FIGS. In this example, there are node groups 60 and 70 between the terminal A10 and the terminal B20. The node group 60 is provided with nodes having communication cards # 0, # 1 and # 2, respectively. 70 is provided with nodes having communication cards # 3, # 4, and # 5, respectively. 9 and 10, the packet addressed to the terminal 20 received from the terminal 10 is received by the card # 0 of the node 60, and the card # 0 does not have a learning result for the terminal 20, and therefore is flooded (step P11). Cards # 1 and # 2 that have received the flooded packet have a common trunk, from which one representative port is selected by a hardware-independent calculation method, and this representative port is applied to Further, it is transmitted (step P12).

  The route having the representative port leads to the card # 3 of the node 70, and even this card # 3 has no learning result regarding the terminal 20 that is the destination. For this reason, the packet is also flooded in the card # 3 (step P13). In this way, the packet reaches the destination terminal 20 through the card # 5. In addition, during the packet transfer from the terminal 10 to the terminal 20 shown in FIGS. 9 and 10, each of the communication cards # 0 to # 5 of each of the node groups 60 and 70, as described with reference to FIGS. Learn information about the route to your machine in association with the source address.

  Next, as shown in FIGS. 11 and 12, it is assumed that a packet is transmitted from the terminal 20 to the terminal 10 this time. In the card # 5 of the node group 70 that receives this, the route to the destination terminal 10 is learned during the transfer operation described with reference to FIGS. 9 and 10, and the route to the terminal 10 as a result of the learning has a trunk. Recognizing this, a route to be applied is selected from among a plurality of routes of the trunk by a calculation method unique to the hardware of the own device. In this case, the route including the card # 4 is selected. In the packet transfer operation described here with reference to FIG. 10, a route passing through the card # 3, not the card # 4, is applied. However, as described above, when trunking is applied, information on the entire trunk, not individual routes, is learned, and selection of a route included in the same trunk is determined by a calculation method unique to the communication card of the transmission source. For this reason, when the transfer directions are different, another card # 4 in the same trunk may be selected instead of the card # 3 constituting the previously applied route.

Thereafter, the packet passes through the card # 3, reaches the card # 2 of the node group 60 from the card # 3 (steps P14 and P15), and further reaches the terminal 10 via the card # 0 in the node 60 group. To do.
JP 2002-247089 A JP 2002-252625 A

  As described above, when the packet is transferred between the terminal 10 and the terminal 20, the hardware (communication card) for selecting the route in the trunk is different, so that different routes in the same trunk may be selected. In this case, the communication cards constituting the respective paths are engaged only in one-way transfer. Therefore, in these cards, the learning result is not utilized, and flooding is always performed during the packet communication.

  This will be described with reference to FIGS. FIG. 13 shows operations after Steps P11 to P16 shown in FIGS. When a packet addressed to the terminal 20 is received again from the terminal 10 at the card # 0, the route of the card # 1 in the same trunk is selected and applied as in the previous time (step P17). The card # 1 that has received this transfers the packet to the card # 3 from the information related to the trunk (step P18). Upon receiving this, card # 3 searches the learning result for terminal 20 that is the destination of the packet. However, as described above, the packet from the terminal 20 to the terminal 10 does not pass through the card # 3 but passes through the card # 4 (step S15). For this reason, the learning result about the terminal 20 does not exist in the card # 3. For this reason, card # 3 floods the packet again (step P19).

  Further, in FIG. 14, when a packet addressed to the terminal 10 is received again from the terminal 20 at the card # 5, the route of the card # 4 in the same trunk is selected and applied as in the previous time (step P20). The card # 4 that has received the packet transfers the packet to the card # 2 from the information related to the trunk (step P21). Upon receiving this, the card # 2 searches the learning result for the terminal 10 that is the packet destination. However, as described above, the packet from the terminal 10 to the terminal 20 does not pass through the card # 2, but passes through the card # 1 (step S17). For this reason, the learning result about the terminal 10 does not exist in the card # 2. For this reason, card # 2 floods the packet again (step P22).

  Further, in the card # 2 in the uppermost stage in FIG. 14, since only one-way transfer from the terminal 10 to the terminal 20 occurs as described above, the transfer cycle becomes longer than in the case of two-way transfer. This also increases the possibility that the corresponding learning result is deleted by aging.

  In this way, there is a node that continues to be flooded during a series of communications, increasing the possibility of occurrence of a situation in which line traffic increases.

  In view of the above problems, an object of the present invention is to provide a packet transfer system in which route learning is effectively performed even when trunking is applied by link aggregation, and the occurrence of flooding can be minimized. To do.

  According to the present invention, each packet transfer device is configured to generate a learning packet at a predetermined timing under a predetermined condition. In other words, when different routes are selected for the going direction and the returning direction due to trunking or the like, and even when a certain apparatus only performs transfer in one direction, at a predetermined timing. By generating a learning packet, for example, a device that performs transfer in only one direction as described above is also forced to receive a packet in the other direction using the learning packet. As a result, each path appropriately receives bidirectional packets, and bidirectional path learning is ensured. As a result, it is possible to prevent an increase in line traffic due to repeated flooding.

  However, since this learning packet is merely intended for effective route learning, if it occurs frequently frequently, it will cause an increase in line traffic. In order to prevent such an adverse effect, it is necessary to determine the learning packet generation conditions, that is, the generation frequency, the generation opportunity, and the like appropriately.

  Furthermore, even if a series of communication has already been completed, if the learning result is not erased by any time and remains in the buffer, the limited buffer storage capacity cannot be effectively used. Therefore, it is necessary to devise a method of utilizing the aging process so that the end of communication is appropriately determined, and in this case, the corresponding learning result is quickly deleted.

  That is, the learning packet is generated only when the route to the destination of the received packet includes a trunk, and when the learning result becomes the hit target of the destination search, not at the first learning time. It is desirable to generate. That is, it is desirable that a learning packet be generated when a packet whose destination is a learned transmission source is received. At this time, it is desirable to transmit the learning packet to all the transfer devices constituting the trunk. Further, after that, it is desirable to generate a learning packet each time the learning result becomes a hit target for destination search again.

  However, during this time, it is desirable that the aging process is performed periodically. That is, it is desirable to delete from the buffer at the time of the last learning, that is, after the elapse of a predetermined time from either reception of the last packet from the same transmission terminal or reception of the last packet from the same destination terminal. Therefore, when a packet is received again from the same source terminal thereafter, it is handled as the first learning.

  According to the present invention, since the result of route learning can be utilized even when the round-trip communication route is different, the occurrence of flooding can be effectively suppressed, and the traffic on the line can be reduced.

  Embodiments of the present invention will be described below with reference to the drawings.

  FIGS. 15 to 20 are for explaining the embodiment of the present invention, and the basic system configuration assumes the system configuration example described with reference to FIGS. Each node configuring the system is, for example, L2SW.

  In FIG. 15, the route to the destination of the received packet includes a trunk, and the node has received a packet from the same address as the corresponding destination address (DA), and has a source address learning result at that time. If there is a hit target in the destination search (step S3 in FIG. 2), the learning packet is transmitted to all nodes having the trunk. That is, in FIG. 15, a packet is once transferred from the terminal 10 to the terminal 20, and then a reply packet from the terminal 20 to the terminal 10 is received at each node passed by the transfer operation, for example, card # 5. In this case, the card # 5 transmits a learning packet to all nodes having the corresponding trunk, that is, the cards # 3 and # 4. As a result, the route for the corresponding MAC address is learned in all these nodes. That is, in this case, the path having the card # 5 is learned for the MAC address “B” of the terminal 20 in each of the cards # 3 and # 4.

  For convenience of explanation, the route of the cards # 1 and # 3 is selected as the trunk representative port in the direction in which the terminal 20 receives, and the card # as the representative port of the same trunk is different in the direction of transmission to the terminal 10. Assume that the route of # 4 and # 2 is selected. In this case, if the learning packet is not generated, flooding repeatedly occurs as described with reference to FIGS. That is, when a packet in the transmission direction for the terminal 20 arrives at the card # 3 again, the card # 3 has not received a packet in the reception direction from the terminal 20, and therefore does not have the corresponding learning result and is flooded again. Become. However, according to the embodiment of the present invention, since the learning packet is transmitted from the card # 5 to the cards # 3 and # 4 as described above, both the cards # 3 and # 4 have the MAC address “B” of the terminal 20. Have learned about. Therefore, even when a packet in the transmission direction to the terminal 20 arrives at the card # 3 again, the card # 3 can specify the route to the terminal 20 according to the learning result by the learning packet.

  FIG. 16 shows an operation time chart of each card in the present embodiment. In the figure, (a) is a clock pulse for performing an aging process. Of the clock pulses, a pulse surrounded by a broken line is a pulse for starting an actual aging process. That is, the aging process is started once every four clock pulse generations. (B) shows the count value of the card LA counter. The counter is updated so as to sequentially take values of 0 to 7 repeatedly as shown in the figure for each clock pulse (a).

  FIG. 16, (c) shows the count value of the content counter. A content counter is provided for each learning result. A vertical broken line indicates the reception timing of the received packet. As shown in the figure, new learning is performed when a packet for a certain source address (SA) is received, and as a result, the value of the content counter is set to “0”. When the next packet for the same MAC address is received, the value of the card LA counter at the reception timing, in this case, “4” is captured. In this case, the learning packet is generated (indicated by an asterisk in the figure) and transmitted to all nodes of the corresponding trunk.

  FIG. 16, (d) shows the state of the learning packet transmitted state flag. The flag is provided for each learning result, and the value “1” is set when a learning packet is generated for the corresponding MAC address. Then, when the packet of the corresponding destination is received again immediately after the learning packet is generated, if the packet is within the period in which the value of the card LA counter is the same “4”, the learning packet is not generated again. That is, the shortest generation period of the learning packet is 1/4 of the aging process start period.

  After that, when the aging process is started, the state flag is cleared to “0” as shown in the figure. Then, as shown in the figure, when the corresponding received packet is received again, the value of the card LA counter has been updated since the previous learning packet generation, so that the learning packet is generated again. As a result, the status flag (d) returns to “1” again as shown.

  After that, when the next aging process is started without receiving the corresponding received packet, the status flag is initially cleared to “0” as in the hit bits shown in steps S11 and S12 of FIG. After that, when the corresponding received packet is not received until the next aging process is activated, the learning result for the corresponding MAC address is deleted as shown in the figure. In FIG. 16, the state where the number is not entered in the content counter in (c) means that there is no learning result.

  Thus, in the case of the configuration shown in FIG. 16 according to the embodiment of the present invention, the learning packet generation interval, that is, the transmission interval is set shorter than the aging process activation interval. That is, learning is performed despite communication continuing due to variations in aging processing timing for each node, fluctuations in processing timing caused by time lag of transmission processing, packet deletion that occurs when the packet flow rate is high, etc. In order to prevent the information from being deleted in the aging process, it is desirable to transmit the learning packet a plurality of times and set the transmission period to be shorter than the aging process activation interval.

  In FIG. 15, when the IP packet addressed to the terminal 20 is received from the terminal 10 by the card # 0 of the node group 60, the MAC address “A” of the terminal 10 and the MAC address “B” of the terminal 20 at that time are shown. No learning results exist for both. In this case, as shown in FIG. 17, the packet is flooded from the card # 0 (step P31).

  The cards # 1 and # 2 that have received this are assumed to have a trunk B that reaches the cards # 3 and # 4 as a route to the terminal 20. The packet is flooded by each of card # 1 and card # 2 (step P32). However, in this case, it is assumed that the card # 1 is selected as the representative port for the trunk B, and transmission from the card # 2 to the card # 4 in the same trunk B is not performed. In response, the card # 3 of the node group 70 performs flooding in the same manner (step P33).

  The cards # 4 and # 5 that have received this are further flooded (step P34). In this case, the card # 3 facing the card # 1 selected as the representative port, not the card # 4, is selected as the representative port for the trunk B. For this reason, transmission from the card # 4 to the card # 2 in the same trunk B is not performed. Note that, during the series of transfer operations, each of the cards # 3, # 4, and # 5 uses the same trunk as the trunk B as the trunk A for the MAC address “A” (SA) of the source terminal 10. To learn as.

  Thereafter, the card # 5 transmits a learning packet to all nodes having the trunk A at a predetermined timing (step P34). As a result, both the cards # 3 and # 4 having the trunk A learn the route having the card # 5 for the MAC address “B” of the terminal 20.

  Thereafter, as shown in FIG. 18, when the card # 5 receives an IP packet from the terminal 20 to the terminal 10, the MAC address “A” of the terminal 10 is learned when the packet is received in step P33. Then, trunk A which is the learning result is applied. At this time, the route by card # 4 is selected from the routes constituting the trunk A by the calculation method unique to the hardware of the card # 5, and this is applied (step P20). In response to this, the card # 4 identifies the card # 2 as the opposing card from the predetermined trunk information and applies it (step P21).

  The card # 2 that has received this information has already learned about the MAC address “A” at the time of packet reception in the previous step P31, so the card # 0 is identified and applied as the corresponding route (step P37). Then, the card # 0 transfers the IP packet to the terminal 10 (no flooding) based on the learning result when the packet is received from the terminal 10 last time (the top row in FIG. 17). Thereafter, the card # 0 transmits a learning packet to all nodes having the trunk B (step P38). As a result, both the cards # 1 and # 2 having the trunk B learn by overwriting the route having the card # 0 for the MAC address “A” of the terminal 10.

  When aging is activated at this time, since both the hit bit and the learning packet transmitted state flag are set to “1”, they are cleared to “0”. The learning results for the respective MAC addresses “A” and “B” are not deleted at this point.

  Then, when a packet is transmitted from the terminal 10 to the terminal 20 again as shown in FIG. 19, the card # 0 that has received the packet overwrites the learning result for the MAC address “A” of the terminal 10. Also, the card # 0 has learned the trunk B as a route for the corresponding MAC address “B” when the packet was received from the terminal 20 last time (step P37 in FIG. 18). Therefore, the route by the card # 1 is selected from the routes constituting the trunk B by the calculation method unique to the hardware of the card # 0, and this is applied (step P41).

  In this case, the corresponding MAC address “B” is hit in the destination search for the terminal 20, and the route to the terminal 20 includes a trunk, so that ¼ of the aging process activation interval has elapsed since the last learning packet generation. If so, the conditions for learning packet generation are satisfied. In that case, a learning packet is transmitted to the cards # 1 and # 2 (step P41). As a result, each of the cards # 1 and # 2 overwrites the learning result for the MAC address “A”.

  Further, when a packet issued from the terminal 10 and passed through the card # 0 reaches the card # 1, the card # 1 identifies the card # 3 as a facing card from a predetermined trunk information and applies this (step) P42). As a result, the card # 3 that has received the packet has already learned about the MAC address “B” at the time of receiving the learning packet in the previous step P34. P43).

  If no learning packet is generated in step S34, flooding is performed here. However, even in that case, if a packet is transmitted from the terminal 20 to the terminal 10 again at the card # 5 as will be described later, the learning packet generation condition is satisfied. A learning packet is sent to learn about the terminal 20 (step P45 in FIG. 20). For this reason, after that, it becomes possible to specify the route and no flooding is required.

  Then, the card # 5 that has received the packet from the terminal 10 in this way forwards the packet to the terminal 20 (no flooding) according to the learning result (the top row in FIG. 18) when the packet was received from the terminal 20 last time. ). Thereafter, the card # 5 transmits a learning packet to all nodes having the trunk A (step P44). As a result, both the cards # 3 and # 4 having the trunk A learn to overwrite the route having the card # 5 for the MAC address “B” of the terminal 20.

  At this point, that is, when aging is activated at the top of FIG. 20, both the hit bit and learning packet transmission status flag are set to “1”, so these are cleared to “0”. The learning results for the MAC addresses “A” and “B” of the terminal 10 and the terminal 20 are not deleted.

  Then, when a packet is transmitted from the terminal 20 to the terminal 10 again as shown in FIG. 20, the card # 5 that has received the packet overwrites the learning result for the MAC address “B” of the terminal 20. Further, the card # 5 has learned to overwrite the trunk A as a route for the corresponding MAC address “A” at the time of the previous packet reception from the terminal 10 (step P43 in FIG. 19). Therefore, the route by the card # 4 is selected from the routes constituting the trunk A by the calculation method unique to the hardware of the card # 5, and this is applied (step P45).

  In this case, in the card # 5, the corresponding MAC address “A” is hit in the destination search for the terminal 10, and the route to the terminal 10 includes a trunk. If / 4 has passed, the learning packet generation condition is satisfied as described above. In this case, a learning packet is transmitted to cards # 3 and # 4 (step P45). As a result, each of the cards # 3 and # 4 overwrites the learning result for the MAC address “B”.

  As described above, when a packet transmitted from the terminal 20 and passed through the card # 5 reaches the card # 4, the card # 4 identifies the card # 2 as a facing card from the predetermined trunk information and applies this card. (Step P46). The card # 2 that has received this has already learned about the MAC address “A” at the time of receiving the learning packet in the previous step P41, so the card # 0 is identified and applied as the corresponding route (step P47). .

  The card # 0 that has received the packet from the terminal 20 in this way forwards the packet to the terminal 10 (no flooding) according to the learning result when the packet was received from the terminal 10 last time (the top row in FIG. 19). ). Thereafter, the card # 0 transmits a learning packet to all nodes having the trunk B (step P48). As a result, both the cards # 1 and # 2 having the trunk B learn to overwrite the route having the card # 0 for the MAC address “A” of the terminal 10.

In addition, this application contains the structure of the additional remarks shown below.
(Appendix 1)
A packet forwarding method for forwarding a received packet by its destination address,
It consists of a learning stage in which each device learns the source address of the received packet in association with its receiving path,
In each device, for each destination address of the received packet, if there is a learning result about the correspondence between the address and the corresponding route, the received packet is transmitted through the corresponding route. It is configured to send the received packet through the route of
A packet transfer method including a learning packet generation step of generating a learning packet for each device at a predetermined timing.
(Appendix 2)
Further, the learning result of the learning step includes a deletion step of deleting at a predetermined deletion cycle,
The method according to claim 1, wherein a period in which the learning packet can be generated in the learning packet generation stage is set to be shorter than the predetermined deletion period.
(Appendix 3)
Each device has a learning packet transmitted flag, and when the learning packet generation stage is executed, the flag is set, and if the flag is set at the execution timing of the deletion stage, it is cleared and the learning result is not deleted. The method according to appendix 2, wherein the learning result is deleted for the first time only when the flag remains cleared at the execution timing of the deletion stage.
(Appendix 4)
The learning packet generation step is executed in a case where the learning packet generation step is executed for a destination address of the received packet that does not have a learning result about a correspondence relationship between the address and the corresponding route. The method described.
(Appendix 5)
The method according to any one of appendices 1 to 4, wherein when applying link aggregation in which a plurality of paths are regarded as a single virtual aggregate path, the learning stage is configured to learn a virtual aggregate path as a reception path. .
(Appendix 6)
In the learning packet generation step, for the destination address of the received packet, the route to the destination is a virtual aggregate route by application of link aggregation, and there is no learning result about the correspondence between the address and the corresponding route. The method according to any one of appendices 1 to 5, wherein a learning packet is generated for all devices having the virtual aggregate path.
(Appendix 7)
A packet forwarding system for forwarding a received packet by its destination address,
Each device that performs packet forwarding has learning means for learning the source address of the received packet in association with the reception path,
In each device, for each destination address of the received packet, if there is a learning result about the correspondence between the address and the corresponding route, the received packet is transmitted through the corresponding route. It is configured to send the received packet through the route of
Each apparatus further includes a learning packet generation means for generating learning packets for other apparatuses at a predetermined timing.
(Appendix 8)
Further, the learning result by the learning means includes a deletion means for deleting at a predetermined deletion cycle,
The packet transfer system according to appendix 7, wherein a period in which the learning packet generation unit can generate the learning packet is set shorter than the predetermined deletion period.
(Appendix 9)
Each device has a learning packet transmission flag, and when the learning packet is generated by the learning packet generator, the flag is set, and when the learning result is deleted by the deletion unit, the flag is cleared. The packet transfer system according to appendix 8, wherein the learning result is not deleted and the learning result is deleted only when the flag remains cleared at the next deletion timing.
(Appendix 10)
The learning packet generation means is configured to generate a learning packet for a destination address of a received packet when the learning packet does not have a learning result about a correspondence relationship between the address and the corresponding route. The packet transfer system according to any one of the above.
(Appendix 11)
The packet according to any one of appendices 7 to 10, wherein when applying link aggregation in which a plurality of routes are regarded as a single virtual aggregate route, the learning unit learns a virtual aggregate route as a reception route. Transfer system.
(Appendix 12)
The learning packet generation means, for the destination address of the received packet, the route to the destination is a virtual aggregate route by application of link aggregation and does not have a learning result about the correspondence between the address and the corresponding route The packet transfer according to any one of appendices 7 to 11, wherein a learning packet is generated for each of the devices, and at that time, a learning packet is generated for all devices having the virtual aggregate route. system.
(Appendix 13)
A packet transfer device for transferring a received packet by its destination address,
Learning means for learning the source address of the received packet in association with the reception path;
If the destination address of the received packet has a learning result about the correspondence relationship between the address and the corresponding route, the received packet is transmitted through the corresponding route. On the other hand, if it does not have the learning result, all the routes are transmitted. It is configured to send the received packet,
Furthermore, a packet transfer apparatus having learning packet generation means for generating a learning packet for another packet transfer apparatus at a predetermined timing.
(Appendix 14)
Further, the learning result by the learning means includes a deletion means for deleting at a predetermined deletion cycle,
14. The packet transfer apparatus according to appendix 13, wherein a period in which the learning packet generation unit can generate a learning packet is set shorter than the predetermined deletion period.
(Appendix 15)
Furthermore, it has a learning packet transmission completed flag, and when the learning packet generating means generates the learning packet, the flag is set, and when the learning result is deleted by the deleting means, the flag is cleared and the learning result is 15. The packet transfer apparatus according to appendix 14, wherein the learning result is deleted only when the flag remains cleared at the next deletion timing without being deleted.
(Appendix 16)
Of the supplementary notes 13 to 15, the learning packet generating means is configured to generate a learning packet for a destination address of the received packet when the learning packet has no learning result about the correspondence between the address and the corresponding route. The packet transfer apparatus according to any one of the above.
(Appendix 17)
The packet according to any one of appendices 13 to 16, wherein when applying link aggregation in which a plurality of routes are regarded as a single virtual aggregate route, the learning unit learns a virtual aggregate route as a reception route. Transfer device.
(Appendix 18)
The learning packet generation means, for the destination address of the received packet, when the route to the destination is a virtual aggregate route by application of link aggregation and does not have a learning result about the correspondence between the address and the corresponding route The learning packet is generated for each of the above, and in that case, the learning packet is generated for all the packet transfer apparatuses having the virtual aggregate route. Packet transfer device.

It is the block diagram (the 1) for demonstrating the route learning function of an example of the past. It is an operation | movement flowchart for demonstrating the example of a conventional route learning function. It is a block diagram (the 2) for demonstrating the route learning function of an example of the past. It is an operation | movement transition diagram for demonstrating the route learning function of an example of the past. It is an operation | movement flowchart for demonstrating the aging processing function of an example of the past. It is a block diagram for demonstrating the aging processing function of an example of the past. It is the operation | movement transition diagram for demonstrating the aging processing function of an example of the past (the 1). It is the operation | movement transition diagram for demonstrating the aging processing function of an example of the past (the 2). It is a block diagram for demonstrating the conventional link aggregation function. It is the operation | movement transition diagram (the 1) for demonstrating the link aggregation function of an example of the past. It is the operation | movement transition diagram for demonstrating the link aggregation function of an example of the past (the 2). It is a block diagram for demonstrating the problem by the repeated flooding generation | occurrence | production by a conventional example. It is the operation | movement transition diagram for demonstrating the link aggregation function of an example of the past (the 3). It is the operation | movement transition diagram for demonstrating the link aggregation function of an example of the past (the 4). It is a block diagram for demonstrating operation | movement of the packet transfer system by embodiment of this invention. It is a time chart for demonstrating operation | movement of the packet transfer apparatus by embodiment of this invention. FIG. 6 is an operation transition diagram (part 1) for explaining the operation of the packet transfer system according to the embodiment of the present invention; FIG. 10 is an operation transition diagram (part 2) for explaining the operation of the packet transfer system according to the embodiment of the present invention; FIG. 10 is an operation transition diagram (part 3) for explaining the operation of the packet transfer system according to the exemplary embodiment of the present invention; FIG. 12 is an operation transition diagram (part 4) for explaining the operation of the packet transfer system according to the embodiment of the present invention;

Explanation of symbols

10, 20 terminal 50-0, 50-1, 50-2 node # 0, # 1, # 2, # 3, # 4, # 5 line card or node 60, 70 node group

Claims (4)

A packet transfer method for transferring a received packet by the destination address,
A learning step of the packet transfer device responsible for packet transfer is learned in correspondence with reception path the source address of the received packet, each packet transfer apparatus per destination address of the received packet, and routes the corresponding with the address of the received packet sent through the route in the case of having a learning result of the correspondence relationship, while transmits the received packet via all routes if no such training results,
Further, the packet transfer apparatus has a learning packet transmission stage in which a learning packet is transmitted at a predetermined timing ,
The learning packet is a packet for causing each packet transfer device to learn the transmission source address of the learning packet in correspondence with the reception path of the learning packet,
In the learning packet transmission stage, each time the packet transfer device receives a packet whose destination is the source address that has already been learned, and the learning-related route includes a virtual aggregate route by application of link aggregation, A packet transfer method comprising transmitting learning packets to all packet transfer apparatuses included in the virtual aggregate route . And a deletion step of deleting the learning result of the learning step at a predetermined deletion cycle,
Cycle may transmit the learning packet in the learning packet transmission step The method of claim 1, wherein Rukoto is shorter than the predetermined deletion period. A packet transfer system for transferring a received packet by the destination address,
Each packet forwarding device responsible for packet forwarding has learning means for learning in correspondence to the source address of the received packet and receiving paths, each packet transfer apparatus per destination address of the received packet, and routes the corresponding with the address correspondence of the case having a learning result of the relationship originated the received packet through the path, while the received packet sent through all the paths when no such training results,
Each packet forwarding device further have a learning packet transmitting means for transmitting a learning packet at a predetermined timing,
The learning packet is a packet for causing each packet transfer device to learn the transmission source address of the learning packet in correspondence with the reception path of the learning packet,
Each time the learning packet transmission means receives a packet whose destination is a transmission source address that has already been learned by the packet transfer device, and each time the learning route includes a virtual aggregate route by application of link aggregation, A packet transfer system, wherein learning packets are transmitted to all packet transfer devices included in the virtual aggregate route . A packet transfer apparatus for transferring a received packet by the destination address,
A learning means for learning the source address of a received packet in correspondence with reception paths, the route if the per destination address of a received packet, has a learning result of the correspondence relation between the path and the appropriate said address the received packet originated, and the other, the received packet sent through all the paths when no such learning result through,
Furthermore, it has a learning packet transmitting means for transmitting a learning packet at a predetermined timing,
The learning packet is a packet for causing each packet transfer device to learn the transmission source address of the learning packet in correspondence with the reception path of the learning packet,
Each time the learning packet transmission means receives a packet whose destination is a transmission source address that has already been learned by the packet transfer device, and each time the learning route includes a virtual aggregate route by application of link aggregation, A packet transfer apparatus, wherein learning packets are transmitted to all packet transfer apparatuses included in the virtual aggregate route .

Images