JPH08251226A - Congestion reporting method - Google Patents

Abstract

PURPOSE: To provide a congestion reporting method with which the congesting state of a repeater node can be speedily and exactly reported to an end node without depending on the flow control system of layer 4 or the protocol class of layer 3. CONSTITUTION: When the generation of sever congestion is detected by a buffer managing part 12, a repeater node congestion abolishing flag CD is set to a received data packet by a layer 4 congestion report control part 15, a layer 4 data part is abolished, and a layer 4 packet setting the packet class to a state reporting packet is generated and transmitted to the destination node. When the generation of light congestion is detected, a layer 4 packet setting a repeater node congestion warning flag CA to the received packet is generated and transmitted to the destination node.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a congestion notification method,
Particularly in an information communication network, the congestion state of a relay node called a gateway or a router that executes a communication protocol process is taken into consideration by a protocol layer 4 of an end node.
The present invention relates to a congestion notification method for notifying to the.

[0002]

2. Description of the Related Art In an information communication network, relay nodes called gateways or routers that execute communication protocol processing are used to interconnect individual LANs constituting the network and other networks or wide area networks. , More value-added services are provided. This kind of relay node performs communication protocol processing using a buffer that temporarily holds the data to be relayed, and if the buffer becomes congested due to a sudden increase in the data to be relayed, Congestion notification for requesting reduction of the amount of transmitted data is given to the end node.

Conventionally, several methods have been proposed as a method of notifying a congestion to an end node. First, as a first conventional technique, there is one that notifies a congestion state using a congestion message (SQ message: Source Quench) as shown in FIG. 10 (for example, TCP / IP).
System Layer Protocol ICMP: Internet Control M
essage Protocol / Document number RFC792, etc.). 10, (a) shows a congestion notification sequence, (b) shows the structure of a congestion message, (c) shows the structure of a data packet, 1 is a relay node, 2 is a source node, 3 is a destination node, 5 Is a communication network that connects the nodes.

In this case, a congestion notification message is sent to the transmitting node 2 in response to the detection of the congestion state at the relay node 1. The ICMP processing unit that monitors the IP processing unit, which is the network layer (hereinafter, referred to as layer 3) of the OSI reference model in the transmission node 2, responds to the congestion message by the transport layer (hereinafter, referred to as the transport layer) of the OSI reference model. The relay node 1 is notified to the TCP processing unit (which is referred to as layer 4) that the relay node 1 is in a congestion state, and T
The CP processing unit reduces the window size, that is, the transmission buffer size to reduce the amount of transmission data.

As a method similar to this, there is a method using a "waiting box" (for example, layer protocol ICMP of TCP / IP system / document number RFC1016). This is to provide a "waiting box" in the IP processing unit of layer 3 so that the transmission data packet from the TCP processing unit of layer 4 waits for a certain time before transmission, and the transmission data is adjusted by adjusting this waiting time. The amount is adjusted. In this case, the ICMP processing unit of the transmission node 2 notifies the IP processing unit of the congestion in response to the reception of the congestion message, and based on this, the IP processing unit increases the waiting time to reduce the amount of transmission data. It was.

As a second conventional technique, as shown in FIG. 11, there is one that provides a congestion indicator in the packet header of layer 3 to notify the congestion state (for example, Random E
arly Detection Gateway for Congestion Avoidance, I
EEE / ACM, Trans.Network, Vol.1, No.4, Aug.1993, pp.397
~ Such). In FIG. 11, (a) shows the congestion notification sequence, and (b) shows the layer 3 packet configuration.
In this case, when the congestion state is light in the relay node 1, the congestion indication is set in the packet header and the data packet is relayed to the destination node 3. The destination node 3 receives this and reduces the amount of data transmitted to the source node 2. Further, when the congestion state at the relay node 1 is severe, the data packet is discarded.

Further, as a third conventional technique, as shown in FIG. 12, there is one for a protocol having window type flow control (for example, Japanese Patent Laid-Open No. 5-91).
144 publication). In FIG. 12, (a) shows a congestion notification sequence, and (b) shows a packet configuration based on window type flow control. In this case, as the window size in the packet header of the data packet to be relayed according to the buffer usage status of the relay node 1,
For example, the WS = 20 is changed to the WS = 10, and the destination node 3 that receives the data transmits the data to the transmission node 2 by using the window size as the transmittable data amount.

[0008]

Therefore, in such a conventional congestion control method, according to the first and second conventional techniques (see FIG. 10), the congestion notification from layer 3 to layer 4 is made at each end node. However, even if the layer 4 protocol is the same between both ends, it is highly possible that the layer 3 protocol will differ depending on the section of each relay node. It was difficult to transfer the notification information to the next section and notify it. Also, since congestion notification from layer 3 to layer 4 is required, when performing protocol processing of each layer header of the received data packet in parallel in parallel for the purpose of improving throughput and shortening processing delay time, simultaneous parallel processing There is a problem that the control becomes complicated.

According to the method of using the "waiting box" of the first conventional technique, even if the transmission data stays in the "waiting box" of the layer 3 for a long time, the layer 4 cannot recognize the transmission data. Requests layer 3 to send data packets one after another and activates a timer for confirmation of response.
In Layer 3, since the transmission data stays in the “wait box”, the response confirmation timer times out, and abnormal processing such as retransmission is likely to be activated, resulting in a problem that throughput decreases.

According to the second conventional technique (see FIG. 11), when the congestion state is severe, the data packet responsible for the congestion notification is discarded, so that the relay node 1 notifies the destination node 3 of the congestion state. It becomes impossible to do. As a result, it takes time for the destination node 3 to detect the occurrence of congestion in the relay node 1, the reduction of the amount of data transmitted from the destination node 3 to the source node 2 in the opposite direction is delayed, and the congestion state of the relay node 1 is reduced. There was a problem that it got worse. Further, due to the discard of the data packet due to the heavy congestion state, it becomes impossible to transmit the control information stored in the packet header to the destination node 3,
In the case of a protocol in which the header of a data packet includes response information for data transmission in the opposite direction, the response information does not reach, and there is a problem that buffer release at each end node is delayed.

Further, according to the third conventional technique (see FIG. 12), since the congestion status of the relay node 1 is notified by the window size which is the protocol information at the layer 4 level, the flow control has the window method. In addition to the limitation, since there is no notification from the relay node 1 to the originating node 2, there is a problem that the congestion notification to the originating node 2 is delayed when there are few packets transmitted in the opposite direction. The present invention is for solving such a problem, and notifies the end node of a congestion state of a relay node quickly and accurately without depending on a flow control method of layer 4 or a protocol type of layer 3. It is an object of the present invention to provide a congestion notification method that can be performed.

[0012]

In order to achieve such an object, a congestion notification method according to the present invention provides a relay node with a buffer usage amount and a plurality of predetermined threshold values in response to packet reception. A buffer management unit that detects the degree of congestion in the buffer by comparison, and a layer 4 packet in which congestion information indicating the degree of congestion detected by the buffer management unit is stored in a layer 4 protocol header When the occurrence of congestion is detected by the buffer management unit, the layer 4 congestion notification control unit outputs congestion information indicating the degree of the congestion state to the layer 4 protocol header unit. The stored layer 4 packet is generated and transmitted to the destination node or the transmission node of the received packet.

Further, the layer 4 protocol header is provided with a relay node congestion discard flag as congestion information indicating the occurrence of severe congestion at the relay node, and the occurrence of severe congestion is detected by the buffer management unit. In this case, the layer 4 congestion notification control unit sets the relay node congestion discard flag of the received data packet, discards the layer 4 data part of the data packet, and sets the packet type to the state notification packet. Is generated and transmitted to the destination node.

Further, a relay node congestion warning flag is provided in the layer 4 protocol header as congestion information, which indicates a slight congestion occurrence in which packets are not discarded at the relay node,
When the buffer management unit detects a slight congestion occurrence, the Layer 4 congestion notification control unit generates a Layer 4 packet in which the relay node congestion warning flag of the received data packet is set and transmits it to the destination node. It was done like this. Further, when the occurrence of congestion is detected by the buffer management unit, the layer 4 congestion notification control unit sets the packet type in which the relay node congestion warning flag or the relay node congestion discard flag corresponding to the detected congestion state is set. A layer 4 packet, which is a notification packet, is generated and transmitted to the source node.

[0015]

Therefore, the buffer management unit of the relay node compares the amount of use of the buffer with a plurality of predetermined thresholds in response to the reception of the packet, detects the degree of congestion that has occurred in the buffer, and detects the congestion. In this case, the layer 4 congestion notification control unit generates a layer 4 packet in which the congestion information indicating the degree of the congestion state is stored in the layer 4 protocol header part and is transmitted to the destination node or the originator node of the received packet. To be done.

[0016]

Next, the present invention will be described with reference to the drawings. FIG. 1 is a block diagram of a relay node according to a congestion notification method according to an embodiment of the present invention. In FIG. 1, 1 is a relay node and 6 is a DSU (line termination unit: Digital Service) connected to each communication line 7. Unit). In the relay node 1, 11 is a buffer for temporarily storing received packets, and 12 is a buffer management unit for detecting a plurality of congestion levels based on the remaining amount of the buffer 11.

Reference numeral 13 is a line control unit for controlling the communication line 7 accommodated via the DSU 6, and 14 is a data link layer (hereinafter referred to as layer 2) and layer 3 of the OSI reference model.
A relay communication processing unit that performs a relay process for the received packet based on the above protocol, and a layer 4 congestion notification control unit 15 that generates a layer 4 packet for notifying the level of congestion that has occurred in the buffer 11. FIG. 2 is a block diagram showing a packet header of a layer 4 packet generated by the layer 4 congestion notification control unit 15. The packet header of a general layer 4 packet shows the congestion level of the congestion occurring in the relay node 1. A relay node congestion warning flag CA and a relay node congestion discard flag CD that are set accordingly are added as congestion information.

FIG. 3 is an explanatory diagram showing the relationship between the remaining amount of the buffer 11 and the congestion level. In this case, the received packets are stored in the buffer 11 in order from the lower side to the upper side as shown in the usage direction. When the usage amount of the buffer 11 reaches a predetermined threshold capacity, the buffer management unit 12 determines that the congestion level corresponds to the threshold capacity. Here, when the threshold capacity 31 is reached, it is detected as a congestion level "1" for prompting a warning, and when the threshold value 32 is reached, it is detected as a congestion level "2" indicating discard.

Next, the operation of the present invention will be described with reference to FIGS. 4 is a flowchart showing the operation of the relay communication processing unit 14, and FIG. 5 is a layer 4 congestion notification control unit 15
3 is a flowchart showing the operation of FIG. First, the line control unit 13 detects the start of communication based on the data packet received via the communication line 7 and the DSU 6, makes a buffer reservation request to the buffer management unit 12, and is allocated according to this request. The received packet is stored in a predetermined area of the buffer 11. After the storage, the relay communication processing unit 14 is notified that the packet has been received.

In response to this, the relay communication processing unit 14 starts the processing shown in FIG. First, the type of activation is determined (step 41). As a result of this determination, the line control unit 13
In the case of the activation by the packet reception notification from the buffer management unit 12, the buffer management unit 12 determines the buffer based on the congestion level detected by comparing the usage amount of the buffer 11 shown in FIG. The state of 11 is checked (step 42). Here, when the buffer 11 is not in the congestion state, it is determined that the congestion control is unnecessary, the transmission line is determined from the destination information of the layer 3 header (step 44), and the corresponding line control unit 13 is notified. Make a request to send a received packet.

On the other hand, when the buffer 11 is in a congestion state, it is determined that congestion control is necessary, and a congestion notification instruction including a congestion level is output to the layer 4 congestion notification control unit 15 (step 43), the process ends. In response to this, the layer 4 congestion notification control unit 15 starts the processing shown in FIG. First, the congestion level included in the congestion notification instruction from the relay communication processing unit 14 is checked (step 51).

Here, when the congestion level of the buffer 11 is the congestion level "1", the relay node congestion warning flag CA in the layer 4 header is set to "1" (= valid), and the relay communication processing unit is set. The transmission request "1" is output to 14 (step 52). Subsequently, the layer 4 congestion notification control unit 15 generates a status notification packet addressed to the transmission node 2 as
A layer 4 status notification packet CT with the relay node congestion warning flag CA set to "1" is generated, and a transmission request "2" is output to the relay communication processing unit 14 (step 5).
3), the process ends.

On the other hand, in step 51, the buffer 1
If the congestion level of 1 is the congestion level "2", the relay node congestion discard flag CD in the layer 4 header is set to "1" (= valid) (step 54), and the packet type of the received packet is confirmed. (Step 55). Here, when the packet type indicates data “DT”, the packet type is changed to the status notification packet “CT”, the data portion is deleted, and the transmission request “1” is issued to the relay communication processing unit 14. Output (step 56).

After sending the transmission request "1", and step 5
If the packet type is other than the data "DT" in 5, the layer 4 state notification packet CT with the relay node congestion discard flag CD set to "1" is generated as the state notification packet addressed to the originating node 2. The transmission request “2” is output to the relay communication processing unit 14 (step 57), and the process is ended. In this way, when the congestion state is severe (congestion level "2"), the buffer of the relay node runs out, so the data packet is discarded, but if the header information part is also discarded, the end congestion state occurs. Therefore, in the present invention, the state notification packet is generated and notified. Since the status notification packet has a shorter length than the data packet, the buffer usage amount is small.

In this way, various transmission requests are output from the layer 4 according to the congestion level, the relay communication processing section 14 is activated again, and the processing shown in FIG. 4 is started. In step 41, the activation type is the layer 4 congestion notification control unit 1
When it is determined that the transmission request from 5 is “1”,
In step S44, the transmission line is determined from the destination information in the layer 3 header, and in step S45, a transmission request is made to the corresponding line control unit 13. As a result, the state notification packet CT generated by the layer 4 congestion notification control unit 15 is transmitted to the destination node 3.

When it is determined in step 41 that the activation type is the transmission request "2" from the layer 4 congestion notification control unit 15, the layer 3 and the layer 2 in which the source node 2 is stored as the destination information. Header is generated (step 46), the process proceeds to step 44, the transmission line is determined from the destination information of the layer 3 header, and a transmission request is made to the corresponding line control unit 13 in step 45. As a result, the state notification packet CT generated by the layer 4 congestion notification control unit 15 is transmitted to the transmitting node 2.

FIG. 6 is a sequence diagram showing a congestion notification method in the case of the congestion level "1". When the level "1" congestion occurs in the relay node 1, the relay node congestion of the data packet DT received thereafter is shown. Warning flag CA is "1"
Is set to the destination node 3 and is transmitted to the destination node 3, whereby the amount of data transmitted in the opposite direction, that is, from the destination node 3 to the source node 2 is reduced.

FIG. 7 is a sequence diagram showing another congestion notification method when the congestion level is "1". When congestion of level "1" occurs in the relay node 1, the data packet DT received thereafter is Relay node congestion warning flag CA
Is set to "1" and is transmitted to the destination node 3, which reduces the amount of transmission data in the opposite direction. In addition, the state notification packet CT in which the relay node congestion warning flag CA is set to "1" is generated and is transmitted to the transmission node 2, which reduces the amount of transmission data to the destination node 3. .

Further, FIG. 8 is a sequence diagram showing a congestion notification method when the congestion level is "2".
When congestion of level "2" occurs at the relay node congestion discard flag CD of the data packet DT received thereafter, the data portion is discarded and the packet type is changed to "CT". Status notification packet C
T is generated and is transmitted to the destination node 3, which reduces the amount of transmission data in the opposite direction.

FIG. 9 is a sequence diagram showing another congestion notification method when the congestion level is "2". When congestion of level "2" occurs in the relay node 1, the data packet DT received thereafter is Relay node congestion discard flag CD
Is set to "1", the data part is discarded, the packet type is changed to "CT", and the state notification packet CT is generated and transmitted to the destination node 3, which causes transmission in the opposite direction. The amount of data is reduced.
Further, the state notification packet CT in which the relay node congestion discard flag CD is set to "1" is generated and sent to the transmission node 2, whereby the destination node 3
The amount of data to be sent to is reduced.

As described above, the buffer management unit 12 of the relay node 1 detects the degree of congestion by comparing the amount of use of the buffer 11 with a plurality of thresholds, and according to occurrence of congestion, layer 4 congestion is detected. In the notification control unit 15, the relay node 1
Since the predetermined layer 4 packet in which the congestion information indicating the degree of the congestion state generated in step 4 is stored in the protocol header of the layer 4 is transmitted to the end node,
It does not depend on the type of the layer 3 protocol of the relay node 1 or the flow control method of the end node at the layer 4, and the end node does not receive the notification from the layer 3 but the layer 4
It is possible to directly grasp the congestion state of the relay node, and to notify the end node of the congestion state of the relay node to the end node quickly and accurately. As a result, the transmission data amount is reduced in the end node, the relay node can quickly recover from the congestion state, and it becomes possible to avoid a decrease in the throughput of the network. Particularly, in the B-ISDN which transmits a huge amount of data, etc. It is particularly effective in high-speed communication networks.

Further, a relay node congestion discard flag CD indicating a heavy congestion state is provided in the layer 4 protocol header, and when the heavy congestion occurrence is detected by the buffer management unit 12, the layer 4 congestion notification control is performed. The unit 15 sets the relay node congestion discard flag CD in the received data packet, discards the layer 4 data part, generates a layer 4 packet in which the packet type is set to the status notification packet, and transmits the layer 4 packet to the destination node 3. Therefore, even when receiving a data packet in a severe congestion state,
By using the header portion, the destination node can be surely notified of the congestion state. Further, it becomes possible to relay the control information stored in the header part of the received data packet to the destination node, and it is possible to avoid a failure caused by discarding the control information.

In addition, a relay node congestion warning flag CA indicating a light congestion state is provided in the layer 4 protocol header, and when the buffer management unit 12 detects a light congestion occurrence, a layer 4 congestion notification control is performed. Since the part 15 generates the layer 4 packet in which the relay node congestion warning flag CA is set in the received packet and transmits the layer 4 packet to the destination node 3, the data packet is discarded before the heavy congestion occurs. It becomes possible to control the amount of data, and it becomes possible to perform avoidance control for the occurrence of severe congestion according to the transmission quality required at the end node.

In response to the occurrence of congestion, the layer 4 congestion notification control unit 15 generates a layer 4 state notification packet in which the relay node congestion warning flag CA or the relay node congestion discard flag CD is set according to the congestion level. Since the transmission is performed to the source node 2, even if the amount of data transmitted from the destination node to the source node is small compared to the congestion notification to the source node via the destination node, The congestion state can be notified to the originating node quickly and accurately.

[0035]

As described above, according to the present invention, a buffer management unit that detects the degree of congestion that has occurred in a buffer in response to the reception of a packet, and congestion information that indicates the degree of congestion that has been detected are layered. A relay node is provided with a layer 4 congestion notification control unit that generates a Layer 4 packet stored in the 4 protocol header unit, and the layer 4 congestion notification control unit detects the congestion state in response to detection of congestion occurrence by the buffer management unit. Since the layer 4 packet in which the congestion information indicating the degree of congestion is stored in the layer 4 protocol header is generated and transmitted to the destination node or the source node of the received packet, the type of the layer 3 protocol of the relay node and the end node It does not depend on the flow control method in Layer 4 of the It is possible to directly grasp a congestion state of the node can be notified of the congestion state of relay nodes for quickly and accurately end node. As a result, the transmission data amount is reduced in the end node, the relay node can quickly recover from the congestion state, and it becomes possible to avoid a decrease in the throughput of the network. Particularly, in the B-ISDN which transmits a huge amount of data, etc. It is particularly effective in high-speed communication networks.

Further, in the layer 4 protocol header, a relay node congestion discard flag indicating occurrence of severe congestion in which packets are discarded at the relay node is provided as congestion information, and when the occurrence of severe congestion is detected, The relay node congestion discard flag of the received data packet is set, the layer 4 data part of the data packet is discarded, and a layer 4 packet with its packet type set to the status notification packet is generated and transmitted to the destination node. Because I did
Even when a data packet is received in the severe congestion state, the destination node can be reliably notified of the congestion state by using the header part thereof. Further, it becomes possible to relay the control information stored in the header part of the received data packet to the destination node, and it is possible to avoid a failure caused by discarding the control information.

Further, a relay node congestion warning flag is provided in the layer 4 protocol header as congestion information, which indicates a slight congestion occurrence in which packets are not discarded at the relay node,
When a slight congestion occurrence is detected, a layer 4 packet in which the relay node congestion warning flag of the received data packet is set and generated and transmitted to the destination node, the data packet is discarded. It becomes possible to control the amount of transmission data before the occurrence of the heavy congestion, and it is possible to perform the avoidance control for the occurrence of the heavy congestion according to the transmission quality required at the end node.

Further, when the occurrence of congestion is detected,
Since the layer 4 packet whose packet type is the status notification packet with the relay node congestion warning flag or the relay node congestion discard flag set corresponding to the detected congestion state is generated and transmitted to the source node, Compared with the congestion notification to the source node via IP, even if the amount of data sent from the destination node to the source node is small, the congestion state of the relay node is notified to the source node quickly and accurately. be able to.

[Brief description of drawings]

FIG. 1 is a block diagram of a relay node that is an embodiment of the present invention.

FIG. 2 is a configuration diagram showing a packet header of a layer 4 packet.

FIG. 3 is an explanatory diagram showing a relationship between a remaining amount of a buffer and a congestion level.

FIG. 4 is a flowchart showing an operation of a relay communication processing unit.

FIG. 5 is a flowchart showing an operation of a layer 4 congestion notification control unit.

FIG. 6 is a sequence diagram showing a congestion notification method (congestion level “1”).

FIG. 7 is another sequence diagram showing a congestion notification method (congestion level “1”).

FIG. 8 is another sequence diagram showing a congestion notification method (congestion level “2”).

FIG. 9 is another sequence diagram showing the congestion notification method (congestion level “2”).

FIG. 10 is a sequence diagram showing a conventional congestion notification method.

FIG. 11 is a sequence diagram showing another conventional congestion notification method.

FIG. 12 is a sequence diagram showing another conventional congestion notification method.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Relay node, 11 ... Buffer, 12 ... Buffer management part, 13 ... Line control part, 14 ... Relay communication processing part, 15 ...
Layer 4 congestion notification control unit, 2 ... Originating node, 3 ... Destination node, 5 ... Communication network, 6 ... DSU, 7 ... Communication line, CA ...
Relay node congestion warning flag (congestion information), CD ... Relay node congestion discard flag (congestion information), 31 ... Threshold capacity (congestion level "1"), 32 ... Threshold capacity (congestion level "2").

Claims (4)

[Claims] 1. An information communication having a relay node for temporarily storing a packet received from a predetermined transmission path in a buffer and transferring it to another transmission path, and performing information communication between end nodes via the relay node. In the network, a buffer management unit that detects the degree of congestion in the buffer by comparing the amount of use of the buffer with a plurality of predetermined thresholds in response to packet reception in the relay node, and the buffer management unit. When a congestion occurrence is detected by the buffer management unit by providing a Layer 4 congestion notification control unit that generates a Layer 4 packet in which congestion information indicating the degree of congestion detected by the Is
The layer 4 congestion notification control unit generates a layer 4 packet in which congestion information indicating the degree of the congestion state is stored in the layer 4 protocol header part, and transmits it to the destination node or the source node of the received packet. Congestion notification method featuring. 2. The congestion notification method according to claim 1, further comprising: in the layer 4 protocol header, a relay node congestion discard flag indicating occurrence of severe congestion in which a packet is discarded at the relay node as the congestion information, When the occurrence of severe congestion is detected by the management unit, the layer 4 congestion notification control unit sets the relay node congestion discard flag of the received data packet and discards the layer 4 data unit of the data packet. Layer 4 with packet type set to status notification packet
A congestion notification method characterized in that a packet is generated and transmitted to a destination node. 3. The congestion notification method according to claim 1, wherein a layer 4 protocol header is provided with a relay node congestion warning flag as the congestion information indicating a slight congestion occurrence in which a packet is not discarded at the relay node, and buffer management is performed. When a slight congestion occurrence is detected by the unit, the layer 4 congestion notification control unit generates a layer 4 packet in which the relay node congestion warning flag of the received data packet is set and transmits it to the destination node. A congestion notification method characterized by the above. 4. The congestion notification method according to claim 2 or 3, wherein when congestion occurrence is detected by the buffer management unit,
The layer 4 congestion notification control unit generates a layer 4 packet in which a packet type with a relay node congestion warning flag or a relay node congestion discard flag set corresponding to the detected congestion state is a state notification packet, and transmits it to the source node. A congestion notification method characterized in that