JP2003023448A - Device for packet abortion - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a device for packet abortion, that is installed in a communication network and that impartially distributes an extra band among connections normally (in a non-congested state), while warranting a minimum contracted band even in a congested state for each connection established in the communication network so as to more efficiently utilize a network resource, economically realizes it, even in a high-speed channel, and is installed in a node so as to realize a simple hardware configuration, thereby realizing congestion avoidance in a node, with high efficiency. SOLUTION: The packet abort device provides a delay for a prescribed time to packets of a connection the reception of which is decided to be rejected.

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention is accommodated in a communication network.
The present invention relates to a device that guarantees the minimum bandwidth of each connection. [0002] In recent years, the rapid growth of the Internet and LAN
Due to widespread use, data traffic such as IP traffic
It is increasing in a number function. Network with this
Congestion occurrence frequency is increasing
Deterioration of screw quality is a problem. For example, conventional
Best Effort that does not guarantee transfer quality on the Internet
Service was mainstream, but best effort service
In some cases, sufficient throughput cannot be obtained with just
ISPs (Internet Service Providers)
When connecting between businesses via high-speed lines, the minimum bandwidth and delay
The demand for services that guarantees extended quality will increase in the future
It is expected to increase. [0003] Guaranteed minimum bandwidth contracted for each user
Services include GFR (Guaranteed Frame Rate) services.
Service. Users can use each other during non-congestion
Bandwidth (remaining bandwidth) can be shared and used. [0004] Regarding the method for realizing the GFR service,
Many proposals have been made, one of which is WRR (Wei
ghted Round Robin) method. FIG. 16 shows the WRR method.
It is a figure showing an outline. The WRR method uses a key for each connection.
Has a queue, weighted for each connection,
By performing read control according to the weight for each queue,
Distribute bandwidth between connections. [0005] All connectors sharing a band are assigned to the WRR.
By accommodating the connection, the weight of the accommodation connection
Band can be distributed according to. According to WRR method
The surplus bandwidth was set to a certain rule while guaranteeing the minimum bandwidth.
It is possible to distribute fairly among connections
You. However, the hardware of the WRR system is complicated.
The speed is proportional to the number of connections to accommodate
It becomes a neck. Also, it is necessary to have a queue for each connection.
Therefore, the amount of hardware in the buffer unit also becomes a problem. One
In short, the WRR method has become faster and accommodated connections.
As the number of services increases, GFR services will be realized economically
It is difficult. Therefore, a simple hardware configuration
There is a need for a method for realizing a GFR service. [0007] GFR service with simple hardware configuration
FIFO-Tagging method
It has been known. In this scheme, the buffer is connected
You do not need to have one for each line, just one for each line. FI
In the FO-Tagging method, the network is entered for each connection.
Observes the input rate to the network at the entrance, and the measured rate
If it is below MCR (Minimum Cell Rate),
Packet passes through as it is and exceeds the MCR.
Then, Tag is added to the header of the packet. here
So, MCR means that the network guarantees the transfer for the connection.
Bandwidth. [0008] A threshold value is provided in the FIFO buffer.
And always monitors whether the queue length exceeds the threshold.
You. If the queue length exceeds the set threshold,
Means that the packet with Tag in the header is FI
Discarded before entering the FO buffer,
Is the packet without Tag, that is, the input level to the network.
Only packets with connections below the MCR are passed.
You. [0009] In the FIFO-Tagging method, the FI
MC of the connection that accommodates the output speed of the FO buffer
R is not less than the sum of R to maintain MCR during congestion.
Proof is possible. Also, when there is no congestion, M
If there is a surplus band that exceeds the sum of CR
, The bandwidth is shared by multiple connections. [0010] However, in this method,
Since multiple connections share a buffer, each connection
Of the band exceeding the sum of the MCR of the
Is it proportional to the input rate to the FIFO buffer?
Is distributed among each connection, so "fairness"
There was a problem in that. FIFO-Tagging
In the formula, if a surplus band occurs,
Because there is no provision on whether to distribute between connections,
In some cases, one connection occupies the surplus bandwidth.
There was a problem that an unfair situation occurred. The present invention has been made in such a background.
However, by installing it in a communication network,
For each connected connection, even during congestion
Normal time (non-congestion) while guaranteeing the contracted minimum bandwidth
Time), the surplus bandwidth is distributed fairly among the connections,
A network that enables more efficient use of network resources
It is an object to provide a device for discarding a ket. The present invention
Is economically feasible for packet discard even on high-speed lines.
It is intended to provide a disposal device. In addition, the packet
Easy installation by installing in a node such as an exchange switch
Hardware configuration and efficient intra-node traffic.
Provide a packet discarding device that can achieve congestion avoidance
The purpose is to do. SUMMARY OF THE INVENTION The present invention provides an
Means for determining whether or not a packet can be accepted, and accepting by this determination
Means to discard rejected packets
It is a packet discarding device provided. The features of the present invention are as follows.
The discarding means determines that the determining means is rejected.
A delay of a certain amount of time for packets on a connection
There is a means. [0014] Thereby, the packet must be discarded.
RTT (Round Trip Time) also increases, and packet transmission
The effect of lowering the sending rate for the source can be expected
Traffic compared to when packets are discarded.
Can be used effectively. The packet is
The same applies to cells that are fixed-length packets.
Can be. [0015] Thereby, by installing in a communication network,
For each connection accommodated in the communication network,
Even at times of congestion, while guaranteeing the contracted minimum bandwidth,
At all times (when there is no congestion), excess bandwidth is
Distributing them more evenly and enabling more efficient use of network resources
That can be economically realized even on high-speed lines.
A waste disposal device can be realized. In addition, the packet
By installing the switch inside a node such as a switch,
Realizes a simple hardware configuration and enables efficient
Congestion avoidance can be realized. (First Embodiment) Packet of the present invention
The multiplexing device of the first embodiment of the present invention to which the discarding device is applied
This will be described with reference to FIGS. Figure 1 shows a multiplexing device
FIG. 2 is a block diagram of a device. Figure 12 shows the monitoring packet
FIG. 9 is a diagram for explaining queue length detection using a packet. The packet discarding device 5 of the present invention is applied.
The multiplexing device, as shown in FIG.
Connection identification for identifying connections 1 to k to which they belong
Unit 1 and packet discard to determine whether the packet can be accepted
Discard device 5 and the packet discard device 5
FIFO buffer 6 for temporarily storing packets
And packets of a plurality of connections are stored in the buffer 6.
Queue of buffer 6 for each connection i
Queue length monitoring unit 7 for detecting length Xi and virtual queue tube
And the weight determined in advance for each connection i.
And a connection information storage unit 2 for holding the value of Wi.
The buffer 6 has a maximum queue length which is a threshold value in advance.
Qmax is set and exceeds the maximum queue length Qmax
Then, the sum of the queue lengths Xi, which is substantially zero, is a series of Xi.
The continuation function is α (ΣXi), and one or more packets are stored in the buffer 6.
The sum of the weights of connection i in which
Wact, and the permitted queue length calculation unit 4
AQi is calculated as AQi = α (ΣXi) · Wi / Wact, and the packet discarding device 5
Is smaller than the permissible queue length AQi.
For packets belonging to the
Set. In this embodiment, a plurality of connections are mixed.
In the situation stored in one buffer 6
The queue length monitoring unit 7 and the queue length monitoring buffer 7
And the virtual queue management unit 3
The detection of the queue length Xi of the f6 is defined as a virtual queue length.
Say. In the following, for the sake of simplicity, the virtual queue
Let the length be simply the queue length Xi or Xj. The packet discarding device 5 checks the connection rejected
The queue length Xj of the action j is the permission queue length AQj.
Calculate the percentage that exceeds, and according to the calculated percentage
To the packet of connection j that was determined to be rejected with probability
Some of the packets are regarded as packets judged to be accepted.
Alternatively, it can be stored in the buffer 6. Further, the packet discarding device 5 determines that the acceptance is rejected.
The packet of the connection j to be determined exists in the buffer 6.
If it is not present, the
It can be determined that the reception is permitted. Arrive at network beyond minimum guaranteed bandwidth
For packets that do, a tag is added to the
The packet discarding device 5 sends the packet to be determined
When a tag is attached, it should be judged that this acceptance is rejected.
Packet is determined to be accepted. As a feature of the present invention, the packet discarding device 5
Indicates that the queue length Xj of the connection
Calculate the ratio exceeding the permitted queue length AQj and reject
The percentage calculated for the packet of connection j determined to be no
After a certain time delay is given,
Stack. The queue length monitoring unit 7 determines the queue length Xi
Queue length Xi is detected for a change cycle less than the change cycle of
I do. Alternatively, the queue length monitoring unit 7
The maximum value of the queue length X as the detection result of the queue length Xi.
Output every unit time. Further, as shown in FIG.
Monitoring packet to send out monitoring packet
An insertion unit 18 is provided, and the queue length monitoring unit 7
Monitoring packet transmitted from the
The time when the packet was input to buffer 6 and the monitoring
Time difference from the time when the packet was output from buffer 6
The queue length Xi can be detected according to the following equation. Also, a plurality of service classes are provided,
In the connection identification unit 1, the connection of the arriving packet is
Service class to which the packet belongs along with the
At the same time, and the virtual queue management unit 3
Queue length for each service class in the buffer 6 for each i
Xi, and the connection information storage unit 2
Service class predetermined for each service i
The packet discarding device 5 holds the value of the weight Wi for each
Calculate the permissible queue length AQi for each service class as AQi = α (ΣXi) · Wi / Wact, and calculate the service class for each connection i.
Queue length Xi for each service class
Packets belonging to connections of length AQi or less
Can be determined to be a reception permission. In the following, the first embodiment of the present invention will be described in more detail.
Will be described. FIG. 14 shows a minimum bandwidth guarantee service to which the present invention is applied.
FIG. 2 is a configuration diagram of a network that provides services. Originating side
The user sends a packet to the destination user through the network.
Is sent. Network is contracted for user
It guarantees the transfer rate up to the minimum bandwidth. Network
Each link has multiple users, and
Determine the area based on the fee and the minimum guaranteed bandwidth.
Distributed among the users according to the weights. As described above, the packet discarding apparatus of the present invention
The multiplexing device provided has a network as shown in FIG.
In the bandwidth control device (UPC: Usage
Parameter Control). Ma
In addition, by providing it in the packet switching switch,
It can also be used to avoid buffer congestion in the switch. First, the calling user sends a packet to the network.
Outgoing port is located at the ingress edge of the network
Packet transmission to user's network by rate observation device
The exit rate is measured. Packet transmission rate must be kept at a minimum.
If it exceeds the authentication band, Tag is added to the header of the packet.
Packets sent below the minimum guaranteed bandwidth.
Is sent out to the network as is. Tag attached
Packets that have not been dropped can be discarded in the network.
Guaranteeing the minimum bandwidth by transferring to the destination user
Do. The multiplexing device of the present invention is located in the relay node,
Link bandwidth shared by several users as weight for each user
Therefore, a process of distributing is performed. As shown in FIG. 1, the multiplexing device
From the service identification unit 1, the bandwidth control unit 10, and the buffer unit 11
It is configured. The connection identification unit 1
Identify the connection from the packet header. Bandwidth control
The unit 10 performs bandwidth control based on the connection information.
Related processing. Buffer section 11 is simple FIFO type
Shared by all users accommodated in the link in the buffer 6
Have been used. A buffer 6 is connected to a queue length monitor 7.
The queue length of the buffer 6 and the packet input / output
The virtual queue manager in the previous stage
3 Where information about input and output packets
Is input to or output from the buffer 6
The packet length of the packet and to which it belongs
Contains the connection identification number. When a packet arrives from the input line side,
The connection identification of the packet is performed by the connection identification unit 1.
Is sent to the subsequent band control unit 10 and enters the buffer 6.
Packets to be discarded and discarded
Is a packet discarded on the spot and not determined to be discarded
Is input to the buffer unit 11 at the subsequent stage and is
Therefore, it is output to the output line. In the present invention, the packet discard in the band control unit 10 is
Bandwidth control based on discarding plays an important role. Accordingly
Next, the processing of the band control unit 10 will be described in detail. band
The area control unit 10 includes a virtual queue management unit 3 and a permitted queue length meter.
Calculation unit 4, connection information storage unit 2, packet discarding device
5. In the virtual queue management unit 3, a queue length monitoring unit
7 and the queue length of buffer 6 based on the information from
Calculate the queue length for each application. The virtual queue manager 3
Connected to the connection information storage unit 2 and calculated
The virtual queue length for each connection
The information is stored in the information storage unit 2. A method for calculating each queue length will be described.
First, regarding the calculation of the queue length of the buffer 6,
The queue length information transmitted from the length monitoring unit 7 is virtually used
It is only stored in the queue management unit 3. Next, connection
This is the method of calculating the virtual queue length for each application.
Since the buffer 6 is empty, the virtual queue of each connection is
-The length is zero. Packet is input to buffer 6
And the packet length and connection identification number of the packet
Is transmitted through the queue length monitoring unit 7. That information
Originally, the virtual queue management unit 3
Access to the virtual queue length of the connection
A process of adding the length of the received packet is performed. packet
At the time of output, on the contrary, the data transmitted from the virtual queue length
A process of subtracting the packet length is performed. FIG. 2 shows an example of the connection information storage unit 2.
Then, for each connection calculated by the virtual queue management unit 3,
Information on virtual queue length, connection weight
Information is stored. Next, the permission queue length calculation unit 4
-Buffer from the management unit 3 and the connection information storage unit 2
Information about the queue length of key 6 and virtual
Receives information about queue length and weight, and
Has a function to calculate the permitted queue length. Here
The entry's permitted queue length is
Indicates the maximum queue length. The total of the permitted queue length AQi of the connection i
The calculation method will be specifically described. Connection weight
Is Wi, and the queue length of the current buffer 6 is Xi.
AQi = α (ΣXi) · Wi / Wact
α (ΣXi) is a continuous function and is a graph of FIGS.
Is an example of a queue length function α (ΣXi). On the horizontal axis
Taking the queue length ΣXi, the vertical axis represents the value of the function α (ΣXi)
Take. Wact is the sum of the active connection weights
It is. An active connection is a virtual queue
Connections with a length greater than zero, at least
A connection where one or more packets are in buffer 6
It is an option. The function α (ΣXi) is a threshold having a queue length.
(Qmax in the figure), it becomes zero. That is,
-If the length exceeds a certain threshold, it is determined that there is congestion.
As a result, the permitted queue length of each connection becomes zero. For example, in the example of FIG.
In the range where no function exists, the function α (ΣXi) takes a constant value,
When the value exceeds the value Qmax, the function α (ΣXi) becomes zero.
Become. That is, the permission key is set within a range not exceeding the threshold value Qmax.
The queue length AQi is distributed in proportion to the weight of the connection.
You. In the example of FIG. 4, the queue length ΣXi is changed from zero.
In the interval up to a predetermined value less than the threshold value Qmax,
The number α (ΣXi) gradually decreases and approaches the threshold value Qmax
The threshold value suddenly decreases as the value increases, and becomes zero at the threshold value Qmax.
That is, the permitted queue length AQi is smaller than the queue length ΣXi.
Is distributed more than the weight of connection i
However, if the queue length ΣXi exceeds a predetermined value,
The distribution rate of the queue length AQi decreases. In the example of FIG. 5, the queue length ΣXi and the function α
The value of (） Xi) is linearly inversely proportional to the value of (ΣXi). That is,
Available queue length AQi is linearly inversely proportional to queue length ΣXi
You. This is the simplest and basic control example. In the example of FIG. 6, the queue length Xi and the function α (Σ
Xi) is inversely proportional to the quadratic function. sand
That is, the permitted queue length AQi is when the queue length Xi is small.
Is large, but suddenly decreases as the queue length Xi increases.
As the queue length Xi approaches the threshold value Qmax
And decrease gradually. This increases the queue length Xi.
At the very beginning, the increase in traffic can be strongly suppressed.
Wear. In the example of FIG. 7, the queue length Xi and the function α (Σ
Xi) is a quadratic function whose value is the inverse of the quadratic function shown in FIG.
It is inversely proportional to the number. That is, the permission queue length A
Qi is large when the queue length Xi is small,
As the length Xi increases, it begins to gradually decrease, and the queue length X
As i approaches the threshold value Qmax, the decrease becomes steeper. This
Accordingly, as the queue length Xi approaches the threshold value Qmax,
As a result, an increase in traffic can be strongly suppressed. In the example of FIG. 8, the queue length Xi and the function α (Σ
Xi) is inversely proportional to the value in a stepwise manner. That is, the permission key
The queue length AQi is large when the queue length Xi is small.
However, as the queue length Xi increases, it gradually begins to decrease,
It suddenly starts to decrease from the predetermined queue length Xi, and the queue length Xi becomes
The decrease gradually decreases as the threshold value Qmax is approached.
You. As a result, the queue length Xi is zero and the threshold value Qmax is
Strongly suppresses traffic increase when near the middle of
be able to. In the example of FIG. 9, the queue length Xi and the function α (Σ
The value of Xi) is inversely proportional to the example of FIG.
You. That is, the permitted queue length AQi is smaller than the queue length Xi.
Large at the time, but the queue length Xi increases
Suddenly begins to decrease and gradually decreases from the predetermined queue length Xi.
At first, as the queue length Xi approaches the threshold Qmax,
Suddenly begin to decrease. This determines whether the queue length Xi is zero.
Between the time when the threshold value starts to increase and when the threshold value Qmax is approached.
The increase in traffic can be strongly suppressed. The queue length Xi is the current queue
Length (the instantaneous value of the queue length when a packet arrives)
However, as the queue length Xi,
An average value may be adopted. The packet discarding device 5 determines the connection between the permitted queue length and the connection.
Packets arriving from the virtual queue length for each
It is determined whether the input is made to the file 6 or discarded. FIG.
The processing flow at the time of packet arrival for deterministic discard processing
It is a flowchart shown in FIG.
If the virtual queue length Xi is equal to or smaller than the permitted queue length AQi,
If the input exceeds the permitted queue length AQi,
Packets with Tag are discarded, and Tag is added.
Packets that have not been dropped are input to buffer 6.
Therefore, while always guaranteeing the minimum bandwidth of each connection,
The occupancy of the buffer 6 is controlled to be fair. The virtual queue length exceeds the permitted queue length.
If the packet has been dropped, the packet is discarded deterministically as described above.
It is also possible to stochastically discard packets instead of
is there. FIG. 11 shows the arrival of a packet for performing a stochastic discard process.
12 is a flowchart showing the processing flow of FIG.
As shown, the allowed queue length is AQi, and the virtual queue length is X
If i, the discard probability P is P = (Xi-AQi) / X
given by i. That is, the virtual queue length is the allowed queue length
If the number exceeds the limit,
You. By controlling the bandwidth by stochastic discard,
When paying attention to the connection, the virtual queue length is permitted
Even when the queue length is exceeded, packets are discarded continuously.
Phenomena are less than when packets are discarded deterministically
Therefore, if the affinity with TCP rate control increases,
I can say. The packet input to the buffer 6 is FIFO
Output to the output line according to the norm. Generation of pseudo random number Rand in FIG.
Is performed every time a packet arrives, and the discard probability P and the pseudo random number R
and discarding based on the result of comparing
Thus, it is possible to actually realize the discard with the discard probability P.
Wear. For example, when the discard probability P is 0.5,
The probability that the random number Rand takes a value larger than 0.5 is also
0.5, and the pseudo-random number Rand is larger than 0.5
By executing the discarding in the case, the discard probability P
Is disposed of. The multiplexer according to the present invention uses a network based on the queue length.
The congestion state of the
The queue length is equal to zero and the rate is below the minimum guaranteed bandwidth.
The network forwards only the packet sent to the network in step (1). Ki
If the queue length is less than a certain threshold, there is an excess band
Each connection is buffered in proportion to the weight.
Occupation allows bandwidth to be distributed fairly among connections
You. A FIFO having a simple hardware configuration is used.
While realizing GFR service with buffer, WRR method etc.
Packet with separate buffers for each connection, such as
In a manner similar to the method of controlling readout of
It is possible to distribute between connections. Also one
Connection is shared by many connections.
Compared to the method that has a buffer for each
The advantage is that the required buffer amount can be
There are points. The function of the packet discarding device 5 is as follows.
If the packet of connection j that should be
If it does not exist in Fa6, it should be judged that this acceptance is rejected
The packet is determined to be accepted. That is, the connection j
The absence of a packet means that connection j
The packet indicates sporadic. like this
Even if sporadic packets are rejected and discarded, continuous
The effect of avoiding congestion is significantly lower than when discarding
Low. Therefore, such sporadic packets
Avoid discarding uselessly. The features of the packet discarding device 5 of the present invention
As a packet of connection j determined to be rejected
After a certain time delay, the data is accumulated in the buffer 6.
During this fixed time, the queue length Xj is equal to the permitted queue length AQj.
Varies according to the ratio exceeded. As a result, the packet must be discarded.
RTT (Round Trip Time) also increases, and packet transmission
The effect of lowering the sending rate for the source can be expected
Traffic compared to when packets are discarded.
Can be used effectively. As a function of the queue length monitoring unit 7, the queue length
Queue length for a change cycle less than or equal to a predetermined change cycle of Xi
Xi is detected. As a result, the instantaneous queue length Xi
Absorbs changes to prevent traffic control malfunctions due to disturbances.
Can be avoided. The function of the queue length monitoring unit 7 is simply
The maximum value of the queue length Xi within the unit time is detected as the queue length Xi.
As a result, output is performed for each unit time. As a result,
Absorbs inter-queue changes in queue length Xi, and
The malfunction of the lock control can be avoided. Further, as shown in FIG.
Monitoring packet to send out monitoring packet
An insertion unit 18 is provided, and the queue length monitoring unit 7
Monitoring packet transmitted from the
The time when the packet was input to buffer 6 and the monitoring
Time difference from the time when the packet was output from buffer 6
, The queue length X is detected. In this way, it is simply based on the number of stored packets.
Compared to the detection of the queue length Xi, the actual delay time
Detect effective queue length Xi including parameters
Can be. Also, packets belonging to one connection
Example of classifying service into two service classes
It is. One service class emphasizes real-time
In the real-time (R) class for important packets
Yes, another service class is real-time
Non-real-time (N
R) Class. In connection identification unit 1 shown in FIG.
To identify the connection of the arriving packet and
Identify the service class of the packet. Also virtual
In the queue management unit 3, packets stored in the buffer 6
Service class information as well as connection information
Get. Thus, the connection information storage unit 2 stores
A table as shown in FIG. 13 can be provided. For example, consider connection # 1
And X belonging to the real-time class as the virtual queue length
1 (R) and X1 belonging to the non-real-time class
(NR) is recorded. In addition, Realta
Im class weight W1 (R) and non-real-time
A lath weight W1 (NR) is recorded. Weight W1
(R) is set heavier than W1 (NR),
Packets belonging to the real-time class
The packet can be read out prior to the packet belonging to the class. With this, real-time characteristics are regarded as important.
Because packet delay can be reduced, users
Can be improved. (Second Embodiment) Packet Discarding Device of the Present Invention
FIG. 15 shows a multiplexing apparatus according to a second embodiment of the present invention to which
It will be described with reference to FIG. FIG. 15 shows multiplexing according to the second embodiment of the present invention.
FIG. 2 is a block diagram of the device. The multiplexing device, as shown in FIG.
Connection 1 to which the cell which is a fixed-length packet
Ｋk, a connection identification unit 1 for identifying
A reception determining unit 15 for determining whether or not the reception is possible, and the reception determining unit 1
FIFO for temporarily storing cells according to the judgment result of 5
Type buffer 6 and the cell arrival rate for each connection i.
A rate detection unit 8 for detecting the connection Ri and each connection i
That holds a value of a predetermined weight Wi
And a cell information storage unit 2 for storing the cell arrival rate Ri.
Is set in advance to a maximum rate that is a threshold,
The arrival rate of the cell, which becomes almost zero when exceeding the large rate
The continuous function of the sum ト Ri of Ri is β (ΣRi), and
One or more cells are stored in the buffer 6 of the buffer 11
Wact is the sum of the weights of connection i
The reachable arrival rate calculator 14 calculates the permitted arrival rate ACRi.
ACRi = β (ΣRi) · Calculated as Wi / Wact
Then, the reception determining unit 15 determines that a cell arrives for each connection i.
A connector whose rate Ri is equal to or less than the permissible arrival rate ACRi.
For cells belonging to the
You. In this embodiment, a plurality of connections are mixed.
In the situation stored in one buffer 6
The queue length monitoring unit 7 and the queue length monitoring buffer 7
And the virtual queue management unit 3
The detection of the queue length Xi of the f6 is defined as a virtual queue length.
Say. FIG. 14 shows a minimum bandwidth guarantee service to which the present invention is applied.
FIG. 2 is a configuration diagram of a network that provides services. Originating side
The user sends a packet to the destination user through the network.
Is sent. Network is contracted for user
It guarantees the transfer rate up to the minimum bandwidth. Network
Each link has multiple users, and
Determine the area based on the fee and the minimum guaranteed bandwidth.
Distributed among the users according to the weights. As described above, the packet discarding apparatus of the present invention
The multiplexing device provided has a network as shown in FIG.
In the bandwidth control device (UPC: Usage
Parameter Control). Ma
In addition, by providing it in the packet switching switch,
It can also be used to avoid buffer congestion in the switch. First, the calling user sends a packet to the network.
Outgoing port is located at the ingress edge of the network
Packet transmission to user's network by rate observation device
The exit rate is measured. Packet transmission rate must be kept at a minimum.
If it exceeds the authentication band, Tag is added to the header of the packet.
Packets sent below the minimum guaranteed bandwidth.
Is sent out to the network as is. Tag attached
Packets that have not been dropped can be discarded in the network.
Guaranteeing the minimum bandwidth by transferring to the destination user
Do. The multiplexing device of the present invention is located in the relay node,
Link bandwidth shared by several users as weight for each user
Therefore, a process of distributing is performed. In the second embodiment of the present invention, the function of the
As a packet of connection j determined to be rejected
After a certain time delay, the data is accumulated in the buffer 6.
During this fixed time, the arrival rate Rj is equal to the permitted arrival rate AC.
It varies according to the ratio exceeding Rj. As a result, the packet must be discarded.
RTT (Round Trip Time) also increases, and packet transmission
The effect of lowering the sending rate for the source can be expected
Traffic compared to when packets are discarded.
Can be used effectively. As described above, according to the present invention,
By installing in a communication network,
Contracts for each connection during congestion.
In normal times (when there is no congestion),
Surplus bandwidth is distributed fairly among connections, making it more efficient
Network resources can be used. Also fast times
A packet discarding device that can be realized economically on the
Can be manifested. Furthermore, packet switching switches, etc.
Simple hardware by installing in a node
Configuration to achieve efficient congestion avoidance in nodes.
Can be

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram of a multiplexer according to a first embodiment of the present invention. FIG. 2 is a diagram showing a table example of a connection information storage unit. FIG. 3 is a diagram showing an example of a function α (ΣXi). FIG. 4 is a diagram showing an example of a function α (ΣXi). FIG. 5 is a diagram showing an example of a function α (ΣXi). FIG. 6 is a diagram showing an example of a function α (ΣXi). FIG. 7 is a diagram showing an example of a function α (ΣXi). FIG. 8 is a diagram showing an example of a function α (ΣXi). FIG. 9 is a diagram showing an example of a function α (ΣXi). FIG. 10 is a flowchart showing a processing flow at the time of arrival of a packet for performing deterministic discard processing. FIG. 11 is a flowchart showing a processing flow when a packet for performing a stochastic discard process arrives; FIG. 12 is a view for explaining queue length detection using a monitoring packet. FIG. 13 is a diagram showing an example of a table of a connection information storage unit according to the first embodiment of the present invention. FIG. 14 is a diagram showing an example of a network configuration to which the present invention has been applied. FIG. 15 is a block diagram of a multiplexer according to a second embodiment of the present invention. FIG. 16 is a diagram for explaining an outline of a conventional WRR system. [Description of Code] 1 Connection identification unit 2 Connection information storage unit 3 Virtual queue management unit 4 Permission queue length calculation unit 5 Packet discarding unit 6 Buffer 7 Queue length monitoring unit 8 Rate detection unit 10 Band control unit 11 Buffer unit 14 Permitted arrival Rate calculation unit 15 Acceptance judgment unit 18 Monitoring packet insertion unit

   ────────────────────────────────────────────────── ─── Continuation of front page    F term (reference) 5K030 GA03 GA13 HA08 HB17 JL07                       KA03 KA13 LA03 LC11 LC15                       MA04 MA13

Claims (1)

Claims: 1. A packet discarding device comprising: means for determining whether or not an incoming packet can be accepted; and means for discarding a packet rejected by the determination. The packet discarding apparatus further comprises means for delaying the packet of the connection determined to be rejected by the determining means for a predetermined time.