JPH06318955A - Packet network and congestion avoiding method therefor - Google Patents

Abstract

PURPOSE:To avoid the congestion of a packet network to prevent abandonment of packets by using an acceleration as the value expressing the degree of increase of the packet communication speed to control the packet communication. CONSTITUTION:A maximum allowable packet communication speed is determined for packet transmission from a transmission packet terminal equipment 3 of the packet network, and the acceleration is provided as the value expressing the degree, in which the packet communication speed can be increased to the allowable packet communication speed, independently of this speed. That is, the future occurrence of congestion is detected by a packet communication speed V(t)=V+alpha.t of a time (t) after where alpha is the upper limit value of the acceleration and V is the present packet communication speed, and it is discriminated whether congestion will occur or not. If congestion will occur based on this discrimination, a congestion prediction signal is transmitted to the transmission packet terminal equipment 3 from a congestion forecasting circuit 25 of a repeating node device 2, and the transmission packet terminal equipment 3 reduces the packet communication speed.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a packet network having variable speed. The present invention uses variable length packets or AT
The present invention relates to a packet network capable of transferring fixed-length packets (cells) such as M (Asynchrous Transfer Mode) at a variable speed, a method for avoiding congestion in the packet network, and a device configuration therefor.

[0002]

2. Description of the Related Art Conventionally, a communication speed on a transmission line, for example, bi
The communication speed expressed in t / sec was a constant reference speed. Also, the replacement equipment was also at a constant reference speed.

By the way, a variable length packet or a fixed length packet (cell) such as ATM is transmitted from a transmitting node to a receiving node via a relay node based on a destination address of the cell, or a line number or a route identification number. In a packet network or an ATM network that transfers data within the network, it is possible to transmit at a timing that is convenient for the user. In this packet network or ATM network, the amount of transfer information per hour can be changed depending on the user's convenience depending on the contract.

The following methods are taken as traffic congestion in such a transmission network and countermeasures against it.

First, in a system in which data transfer in a network is performed by call admission control, when traffic congestion occurs, transmission from each terminal node with a new communication request is disabled. In this case, the communication request is made again after some time, and if the congestion is eliminated, the request is accepted and communication becomes possible. In this case, even if congestion occurs, the terminals that have been communicating before the congestion occur can communicate, but the call admission control is complicated.

Next, in the packet network, even if congestion occurs, the packet can be sent from each terminal node, but the packet overflowed by the congestion is discarded.
As a result, the sending node is notified of the discard and is further retransmitted, which makes it practically impossible to communicate.

For this reason, in the packet network, there is a method of restricting transmission to the transmitting node and stopping the transmission when traffic congestion is detected as a countermeasure against congestion. The principle of congestion avoidance by this method is shown in FIG. This method is a method of stopping the transmission of cells from the transmitting node by the congestion notification from the receiving node (or the relay node).

However, this method has a problem that packets are discarded when the buffer capacity of the terminal node is small. Further, when the buffer provided is small in capacity, there is a problem that the transmission stop time becomes long. In addition, a large capacity buffer is required to operate the network efficiently,
The efficiency of the entire network deteriorates, which causes the cost of the entire system to increase. Also, a method that does not cause congestion by constructing a network with sufficient margin is conceivable. However, constructing a network with a sufficient margin results in providing a device of a size that is not normally used, such as a relay node or a terminal node, with a buffer of sufficient size, resulting in poor efficiency.

Further, as a congestion control method in a packet network, a method of controlling a packet transfer by cutting a short time during packet transfer, monitoring the packet transfer amount, and judging whether or not congestion has occurred There is (this is called window flow control). This technique is disclosed in JP-A-3-174.
848. The principle of congestion avoidance by this method is shown in FIG. With this method, the round trip delay time is measured at the sending node, and if the round trip delay time is decreasing, it is predicted that the network is not in a high load state, so the transmission allowance on the sending side is increased. When the round-trip delay time is increasing, it is predicted that the network is in a high load state, so the method is to stop the packet transmission by reducing the transmission permission amount on the transmission side.

However, in this window flow control,
If the buffer provided in each node has a small capacity, there is a problem that packets are discarded. In addition, the prediction of the network state due to the delay time may not match the actual state.

[0011]

The conventional problems relating to the congestion detection are as follows. Since the conventional congestion detection determines the state of the transmission line at the present time, it is in a congestion state, so to speak. It is detected, and the congestion of the packet communication at the future time is not predicted based on the packet communication speed at the present time to prevent the congestion.

Further, the conventional packet transfer rate detection for congestion detection is based on a distribution function model in order to predict the packet communication rate when packets from a large number of signal sources are multiplexed. The speed was predicted, and it could not be dealt with when the parameter of the signal source changed with the passage of time. Therefore, it is not intended to detect the occurrence of congestion by predicting the packet transfer rate after a predetermined time in the future.

An object of the present invention is to introduce the concept of acceleration or acceleration ratio as a value representing the degree of increase in packet communication speed in avoiding congestion in the packet network, and controlling packet communication in the packet network based on this concept. To provide a packet network in which packet discard does not occur.

It is another object of the present invention to provide a packet network which can achieve high throughput even if the buffer has a small capacity.

[0015]

The first aspect of the present invention is to:
The present invention relates to a packet network that performs congestion avoidance based on a value indicating the degree of increase in packet communication speed, and a packet network in which packets are transmitted and received between a transmitting node and a receiving node via a relay node, In a packet network in which an upper limit is set for the packet communication speed allowed for the transmitting node, an upper limit is set for a value indicating the degree of increase in the packet communication speed of the transmitting node.

As a value representing the degree of increase of the packet communication speed, an acceleration which is a time change rate of the packet communication speed or an acceleration ratio which is a time change rate of the packet communication speed can be used.

A lower limit or a starting speed (initial speed) can be set for the packet communication speed of the transmitting node of this packet network.

The relay node or the receiving node of the packet network is based on the packet communication speed of the passing packet or the packet communication speed of the arriving packet and the upper limit value of the value indicating the degree of increase of the packet communication speed. T
Means may be provided for predicting the packet communication speed in seconds.

The relay node or the receiving node may include means for predicting that congestion will occur based on the predicted packet communication speed.

Further, the relay node or the receiving node can be provided with means for notifying the transmitting node of the congestion when the congestion is predicted.

A second aspect of the present invention relates to the prediction of the packet notification rate and the congestion prediction, and the packet communication rate is increased for each destination address or line number or route identification number of the packet transferred on the transmission line. Is provided in a packet network in which an upper limit is specified for the value indicating the degree of, and a means for detecting the current packet communication speed, the detected packet communication speed, and the increase in the specified packet communication speed. And a means for predicting the maximum packet communication speed after t hours based on the upper limit of the value indicating the degree of.

The speed predicting means for predicting the maximum packet communication speed sets the upper limit of the specified packet acceleration to α _i.
(I = 1 to m), and the current total rate of packets is V _Σ
Then, the maximum packet communication speed V (t) after t hours is

[0023]

[Equation 4] It is possible to provide a means for calculating and obtaining.

Further, the speed prediction means sets the upper limit of the specified packet acceleration ratio to exp (β), divides the transmission line into n groups, and for each group, the maximum acceleration ratio coefficient β _i in the group, From the maximum acceleration ratio e ^βi (i = 1 to n) and the current packet communication speed V _{G1 to} V _{Gn of} each group, the maximum packet communication speed V (t) after t hours is calculated.

[0025]

[Equation 5] However, VI _j : start speed (initial speed) of each transmitting node, S:
It is possible to provide means for calculating by all the transmission nodes (note that there is no second term when a lower limit is set for the packet communication speed of the transmission nodes).

A third aspect of the present invention relates to packet transmission control in a transmitting node based on a congestion prediction signal, in which a packet is transmitted and received via a receiving node and a relay node, and the packet to be transmitted is permitted. In a packet network transmission node having an upper limit in packet communication speed, means for transmitting a packet at a communication speed having a value within the range of the packet upper limit communication speed that is less than or equal to a predetermined value and represents a degree of increase in packet communication speed. , A means for reducing the packet communication speed when receiving a congestion predicting signal indicating that congestion will occur from the relay node or the receiving node.

It is possible to provide a means for reducing the packet communication speed by the arrival of the request for increasing the packet communication speed from the relay node or the receiving node instead of the arrival of the congestion prediction signal when congestion is predicted.

When the congestion prediction signal is received, the packet transfer interval k corresponding to the packet communication speed at that time is received.
It is possible to provide means for transmitting packets at a transfer interval of (k> 1) times. Also, when receiving the congestion prediction signal,
It is possible to provide means for transmitting the packet at a rate that is the packet communication rate reduced at a constant rate or a constant exponential rate than the packet communication rate at that time.

A packet transmission control circuit for reducing the packet communication speed based on the received congestion prediction signal can be provided.

This packet transmission control circuit is provided with a packet transmission interval time table having as an address a time lapse when the packet transmission interval exponentially shortens, and packet transmission can be performed using this table. Further, when the packet transmission control circuit rises to the specified packet communication speed, the packet transmission control circuit returns to the time elapsed address corresponding to the specified packet communication speed, and from there, again uses the packet transmission interval time of the table to transmit the packet. It can be carried out. Further, when the packet transmission control circuit receives the congestion predicting signal from the network, it can return the address from the time lapsed address of the table to a fixed number or an address reduced by a fixed ratio to perform packet transmission.

A fourth aspect of the present invention relates to a congestion avoidance method in a packet network, in which packets are transmitted and received between a transmitting node and a receiving node via a relay node, and the transmitting node is allowed to do so. In the method for avoiding congestion in a packet network in which an upper limit is set for the packet communication speed, the transmitting node has a packet within a range of an allowable packet communication speed and an upper limit of a value indicating a degree of increase of the packet communication speed. The relay node or the receiving node based on the packet communication speed of the passing packet or the packet communication speed of the arriving packet and the upper limit value of the value indicating the degree of increase of the packet communication speed. The packet communication speed after t hours is predicted, and when congestion is predicted, the fact that congestion is predicted is notified to the transmitting node, and the transmission is performed. Over de When receiving the notification that the congestion is foreseen, characterized in that to reduce the packet transmission rate at that time.

The value indicating the degree of increase in the packet communication speed may be acceleration, or may be an acceleration ratio.

When the transmitting node receives the notification that congestion is predicted, the transmitting node can reduce the packet communication speed by a certain value. Further, the packet communication speed can be reduced by a fixed ratio or a fixed index ratio.

Further, the notification that congestion is predicted can be sent by not transmitting the packet communication speed increase request signal to the transmitting node.

A lower limit or an initial rate can be set for the packet communication rate of the transmitting node.

[0036]

According to the present invention, the maximum allowable packet communication speed and the minimum packet communication speed of zero or more are set for sending packets from the transmitting node of the packet network. As a value representing the degree to which the packet communication speed is increased, a speed increase rate, that is, an acceleration, or a speed increase rate, that is, an upper limit of the acceleration rate is provided. Each transmitting node can increase the packet communication speed up to the maximum allowable packet communication speed within this acceleration or acceleration ratio range.

For detection of future congestion occurrence, for example, when the upper limit value of the acceleration is α and the current packet communication speed is V, the packet communication speed after t hours is V (t) = V + α
It can be obtained by the formula of t, and it is possible to judge whether or not congestion is predicted based on the predicted packet communication speed.
Based on this prediction, the relay node or receiving node
When congestion is predicted, a congestion prediction signal is transmitted to the transmitting node.

When the transmitting node receives a congestion predicting signal indicating that congestion is predicted from the relay node or the receiving node of the packet network, the transmitting node reduces the packet communication speed. Instead of the congestion prediction signal, the speed increase request signal display may be deleted to notify the congestion prediction to the transmitting node and reduce the packet communication speed of the transmitting node.

The packet transmission control controls the packet communication speed of the packet to be transmitted to the transmission line at the transmission node when the congestion prediction signal or the speed increase request signal does not arrive. In this packet transmission control, a packet transmission interval time table set so that the packet transmission interval decreases exponentially is provided so that the packet communication speed can be increased exponentially, and this table must be accessed. It is possible to obtain an exponentially increasing packet communication speed at. When the maximum speed is reached, the table access is controlled so as to maintain that speed, and when congestion is predicted, a fixed ratio or a fixed number of address values are returned to the reduced address table. , Control so as to reduce the packet communication speed of the transmitting node.

[0040]

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

FIG. 3 shows an example of a network configuration to which the present invention is applied. Relay node devices 2 interconnected by a transmission line 1 and packets accommodated in each of the relay node devices 2 are shown. It is composed of the terminal device 3.

This packet network is connected to one relay node device 2
Focusing on the above and simplifying the description, it can be simplified to a form in which the transmission packet terminal device and the reception packet terminal device are connected via the relay node device 2 as shown in FIG. Thus, even if the network is simplified to one relay node device 2, the generality of the congestion avoiding operation is not lost. That is, there is no topology restriction such as LAN media access control (eg, Ethernet, token passing, FDDI, DQDB).

(Embodiment 1) An example of controlling packet transfer using the acceleration of the packet transfer speed is shown below.

FIG. 5 shows the configuration of the congestion avoidance method, which is composed of the packet terminal device 3 on the transmission side (hereinafter referred to as the transmission packet terminal device 3) 3 and the relay node device 2 based on the configuration shown in FIG. Show. The configuration of the receiving side terminal device is symmetrical to that of the transmitting side, and therefore its description is omitted.

The relay node device 2 has a buffer memory 21.
And a congestion prediction circuit 25. Further, the transmission packet terminal device 3 has a packet assembling circuit 31, a buffer memory 32, and a packet disassembling circuit 34, and further, a packet flow control circuit 35 for controlling packet transmission from the buffer memory 32, and a transmission from the relay node device 2. The congestion prediction signal receiving circuit 39 which receives the congestion prediction signal thus generated and notifies the packet flow control circuit 35 of the received congestion prediction signal.

The data to be transmitted is the packet assembling circuit 31.
Are packetized and sent to the buffer memory 32. The packets in the buffer memory 32 are controlled by the packet flow control circuit 35 at packet transfer intervals corresponding to a predetermined low-speed packet transfer speed (initial speed), and the packet transmission path 1
Transmission to 1 is started. The packet flow control circuit 35
Until the maximum packet transfer rate is reached over time,
The packet transfer interval is shortened while keeping the packet transfer increase amount per unit time, that is, the acceleration of the packet transfer rate below a prescribed value. When the congestion prediction signal receiving circuit 39 receives the congestion prediction signal,
The packet transfer rate is reduced so that the transfer interval becomes k (k> 1) times the packet transfer interval corresponding to the packet transfer rate at that time, and the packet transfer rate is increased while keeping the acceleration below the specified value again.

Here, an example of a method of calculating the packet transfer rate will be shown. The minimum packet transfer rate (initial speed) of zero or more is v _min (bit / sec), and the maximum packet transfer rate is v _max (bit
/ sec), acceleration α (bit / sec ² ), fixed packet length L
(bit), the offset time corresponding to the minimum packet transfer rate is t ₀ , the packet transfer rate increase time from the minimum packet transfer rate including the offset time to the maximum packet transfer rate is T (sec), and If the current elapsed time including the offset time is t _now (sec), the current packet transfer interval p _now (sec)
And the packet transfer rate v _now (bit / sec) is α = v _max / T (1) p _now = L / (αt _now ) (v _min · T / v _max ≤t _now ≤T) (2) ) p _now = L / v _max (t _now > T) (3) v _now = L / p _now (4). This v _min , v _max , v _now , α, T,
The relationship between t _now and t ₀ is shown in FIG. 6, the horizontal axis is the elapsed time within the packet transfer rate increase time, and the vertical axis is the packet transfer rate.

At the time of receiving the congestion prediction signal, t / k (k>
As 1) the elapsed time of the newly currently t (t is rounded up to t ₀ in the case of t ₀ less), (1) to (4) determine the packet transfer interval p a formula. However, even in this case, t
The maximum value of is T. This makes the current packet transfer rate 1 / k. The method is disclosed in JP-A-3-174848.

In each relay node device 2 of the packet network, the congestion prediction circuit 25 monitors the congestion status of the packet transmission path, and when congestion is predicted by this, packet transmission in the reverse direction as a congestion prediction signal. It is sent out to the transmission packet terminal device using this packet transmission line via the line 11. The number of packets in a unit time may be the number of packets in a sliding window unit time described later. Further, the average value detection method that can be realized by a first-order or higher-order recursive filter may be used.

The threshold value V _th is a time 2 which is equal to or longer than the maximum round trip delay time between the transmission packet terminal device 3 and the relay node device 2.
The product (2D × α × n) obtained by multiplying D by the acceleration α in the transmission packet terminal device 3 and the number n of packet terminal devices that use the packet transmission line 11 is the current packet transfer rate V of the packet transmission line 11 _When 2D × α × n + V _now = V _{max in} which the value including _now is equal to the allowable packet transfer rate V _max , V _th ≦ V _now is set.

FIG. 7 shows the relationship between the packet flow rate per unit time output from the transmission packet terminal device 3 and the packet flow rate per unit time input to the relay node device 2. Here, the transmission delay time between the transmission packet terminal device 3 and the relay node device 2 is D. The relay node device 2 predicts congestion at time b and sends a congestion prediction signal to the transmission packet terminal device 3. The transmission packet terminal device 3 receives the congestion prediction signal at time c after the D time, and reduces the packet transfer rate to 1 / k (k> 1).
In the relay node device 2, after the congestion prediction signal is transmitted, 2
The packet transfer rate decreases at time d after D hours. Therefore, the transmission delay time D and the number n of packet terminal devices are taken into consideration and the upper limit acceleration α, the threshold value, and the congestion prediction signal in the transmission packet terminal device 3 are transmitted so that the buffer overflow does not occur between the time b and the time d. It is necessary to select the minimum period.

The method of multiplying the current packet transfer interval by k based on the congestion predicting signal permits all packet terminal devices that send packets to use the transmission path 11 fairly. The method of decreasing the current packet transfer rate by a constant value by the congestion prediction signal allows the packet terminal device to preferentially use the transmission path 11.

FIG. 8 shows a simulation result of congestion avoidance in the relay node device 10 when k = 1.25. The simulation parameters are as follows.

Fixed packet length L: 53 bytes Packet transmission line speed (packet transfer speed): 150 Mbit / sec Number of packet terminal devices using congestion relay node device n: 100 Minimum packet transfer speed A of transmission packet terminal device (first Speed): 300 Kbit / sec Maximum packet transfer rate of transmission packet terminal device v _max : 15 Mbit / sec Packet transfer speed increase time of transmission packet terminal device T: 250 msec Packet transfer increase rate of transmission packet terminal device (acceleration: α ): 58.8Mbit / sec ² round trip delay time 2D: 2msec threshold value (average packet transfer rate in a sliding window of unit time 1msec) _Vth : 150 × 0.85Mbit / sec Congestion prediction signal minimum period: 4.5msec g in the figure , Input packet amount to the buffer memory 21 of the relay node device 2 observed every 100 packet time (0.282 msec) as a percentage Have the meanings indicated, f is obtained by displaying the number of packets up to Q-length of the buffer memory 21, e is obtained by displaying the time of occurrence of the congestion prediction signal. It can be seen that if the buffer capacity is about 40 packets, packet discard can be avoided and the throughput can be secured at about 0.9.

(Embodiment 2) Next, a configuration for controlling packet transfer using the acceleration rate of the packet transfer rate will be described with reference to FIGS.

Here, the congestion prediction circuit 25 of the relay node device 2 has a time 2D which is equal to or longer than the maximum round trip delay time between the transmission packet terminal device 3 and the relay node device 2 and an acceleration ratio coefficient β (acceleration ratio e ^β ) And the current packet transfer rate V _now of the packet transmission line (2D × β × V _now )
To the value obtained by adding V _now to the above, or V _now × exp (β ・ 2
If D) exceeds the allowable packet transfer rate V _max , it is detected as congestion prediction.

When the congestion prediction signal receiving circuit 39 receives the congestion prediction signal, the packet transfer rate v at that time
A value obtained by dividing _now by the minimum packet transfer rate A (v _now /
A) is multiplied by a constant y (y <1) and is multiplied by the minimum packet transfer rate A, and the packet transfer rate is reduced by setting the packet transfer interval corresponding to a value (A × (V _now / A) ^y ). Let

Here, an example of a method of calculating the packet transfer rate will be shown. Minimum packet transfer speed (initial speed) is A (bit / s)
ec), the maximum packet transfer rate is v _max (bit / sec), the acceleration ratio coefficient is β (1 / sec), and the packet transfer rate increase time from the minimum packet transfer rate to the maximum packet transfer rate is T
(Sec), the fixed packet length is L (bit), and the current elapsed time within the packet transfer rate increasing time is t _now (se
c), the current packet transfer interval p
_Now (sec) and packet transfer rate v _now (bit / sec) are v _now = A x exp (β x t _now ) (5) p _now = L / v _now (6) β = (1 / T) log (v _max / A) (7)

Further, the packet transfer rate V _up (bit / sec) after δt (sec) time from the present time t _now is: v _up = A × exp (β × (t _now + δt)) = v _now · exp (β · δ _t) ≒ v becomes _{now (1 + β · δ t} ) ···· (β · δt«1) = v now + v now × β × δt ... (8). Therefore, the acceleration at this point is v _now × β × δt / δt = v _now × β (bit / sec ² ), which is proportional to the packet transfer rate at the current point.

When the congestion prediction signal is received, y × t _now (y
<1) is newly set as the current elapsed time t _now , and the packet transfer rate v _down (bit / sec) reduced by the equation (5) is calculated, or v _down = A × exp (β × y × t _now ) = It suffices to obtain A × (exp (β × t _now )) ^y = A × (V _now / A) ^y (9). This is equivalent to the constant index ratio reduction method,
The speed may be reduced simply as (v _now ) ^y .

The threshold value V _th is the time 2 which is equal to or longer than the maximum round trip delay time between the transmission packet terminal device 3 and the relay node device 2.
The value obtained by multiplying D by the acceleration ratio coefficient β (acceleration ratio e ^β ) and the current packet transfer rate V _now of the packet transmission line is V _now.
Value (V _now + V _now · β · 2D) or V _now ·
If exp (β · 2D) is equal to the allowable packet transfer rate V _max , that is, V _now = V _max / (1 + β · 2D) (10), V _th ≤V _now Set.

FIG. 9 shows the time relationship between the packet flow rate output from the transmission packet terminal device 3 and the packet flow rate input to the relay node device 2. The transmission delay time between the transmission packet terminal device 3 and the relay node device 2 is D. Note that the transmission delay time D includes delay fluctuations in the relay node device 2. The relay node device 2 predicts congestion at time b of the packet transfer rate corresponding to the threshold, and sends a congestion prediction signal to the transmission packet terminal device 3. The transmission packet terminal device 3 uses the D
The packet transfer rate is reduced D hours after the congestion prediction signal is transmitted at time c after the time, and the packet transfer rate is reduced at the relay node device 2 at time d 2D after the congestion prediction signal is transmitted. Therefore, the transmission delay time D is set so that the buffer does not overflow from the time b to the time d.
In consideration of the congestion prediction time and the transmission packet terminal device 3
, A coefficient β for determining the packet transfer acceleration ratio, a coefficient y (y <1) for reducing the packet transfer rate at the time of receiving the congestion prediction signal, a threshold value V _th , and a minimum transmission cycle of the congestion prediction signal. There is a need.

A method of setting this parameter will be described.

Now, the N transmission packet terminal devices are using the packet transmission line of interest, and the current packet transfer rate of the packet terminal device i (i = 1 to n) is v.
Let _{now, i} be the packet transfer rate V _now observed on the packet transmission line of interest.

[0065]

[Equation 6] Becomes Further, the packet transfer rate V _up predicted on the packet transmission line of interest after Tp time from the present time is

[0066]

[Equation 7] Becomes

In this way, V _up can be predicted without knowing the number of transmission packet terminal devices that are using the packet transmission line of interest. However, the amount of packets transmitted by the transmission packet terminal device immediately after the start of packet transfer cannot be predicted by the relay node device, but can be ignored because the amount of packets is small. If it cannot be ignored, a fixed packet transfer rate may be secured as an offset value and the threshold value V _th may be set. Therefore, if T _p = 2D and V _up = V _max , then (10),
From the equation (12), the threshold value V _th can be determined regardless of the number of terminals in use.

In the case of this embodiment, when the transmitting packet terminal device receives the congestion predicting signal from the relay node device, the higher the packet transfer rate at that time, the larger the packet transfer rate is lowered. Therefore, in the relay node device, if the congestion of the packet transmission path continues for a certain time or longer, the packet transfer rates of the transmitting terminal devices in use become almost equal. Therefore, fairness can be ensured for the transmission packet terminal device in use. If the packet transfer rate is reduced by a fixed ratio when the congestion predicting signal is received, the packet transmitting terminal can be given the priority to use the network. This is the same even if the packet transfer rate is calculated with the new elapsed time obtained by subtracting the constant time t from the current elapsed time t _now in the equation (5).

Here, simulation results of congestion avoidance in the relay node device 2 and the transmission packet terminal device 3 when the deceleration coefficient y = 0.87 are shown in FIGS. The simulation parameters are as follows. However, it is assumed that the round-trip delay time between each transmission packet terminal device 3 and the congestion relay node device 2 is uniformly distributed between 2 msec and 3 msec.

Fixed packet length L: 53 bytes Packet transmission line speed (packet transfer speed): 150 Mbit / sec Number of transmission packet terminal devices that can simultaneously use congestion relay node device n: 100 Minimum packet transfer speed of transmission packet terminal device A (initial rate): 300 Kbit / sec Maximum packet transfer rate of transmission packet terminal device v _max : 15 Mbit / sec Packet transfer rate increase time of transmission packet terminal device T: 100 msec Acceleration ratio coefficient β of transmission packet terminal device: 39.12 (1 / sec) Transmission packet Acceleration ratio of terminal device exp (β): 9.76 × 10 ¹⁶ Round trip delay time 2D: 2 msec to 3 msec Threshold (average packet transfer rate in a sliding window of 1 msec unit time) V _th : 150 × 0.85 Mbit / sec Congestion prediction signal transmission minimum period: 4.5 msec In FIG. 10, a large number of transmission packet terminal devices 3 start packet transfer all at once. In case, congestion relay node packet transfer rate of the apparatus 2 and maximum buffer memory Q
This is the result of observing a long time transition for every 1000 packets (2.82 msec). In the figure, f is the number of transmission packet terminal devices during packet transfer, g is the maximum Q length of the buffer memory within the observation time interval, and e is the packet transfer speed at the buffer input of the congestion relay node device and the packet transfer speed is 15.
It is a percentage display based on 0 Mbit / sec.
h indicates the time of occurrence of the congestion prediction signal. Through this simulation, a large number of transmission packet terminal devices 3
It can be seen that even when all of the packets start to be transmitted all at once, congestion is stably avoided with a small amount of buffer and no buffer overflow.

FIG. 11 shows 100 transmission packet terminal devices 3.
Is using the packet transmission line of the congestion relay node 2, 100 packets (0.282 m
It is the result of observation every sec). In the figure, k represents the maximum value of the sliding window output at the observation time in percent, and e, h, and g are the same as in FIG. By this simulation, a congestion prediction signal is generated when the sliding window output exceeds the threshold,
It can be seen that congestion avoidance is being performed.

(Third Embodiment) Next, a method for avoiding congestion in a packet multiplexing integrated network in which the packet network accommodates a variable speed terminal device and a fixed speed terminal device will be described.

FIG. 12 shows one relay node device 20, one variable speed terminal device 30 and one fixed speed terminal device 40 in the packet multiplexing integrated network. Here, the relay node devices 20 are interconnected in a mesh via the transmission path 11 in the packet multiplexing integrated network. The variable speed terminal device 30 and the fixed speed terminal device 40 are connected to the corresponding relay node devices 20 via the transmission path 11, and are set between them 1: 1, 1: N, N: 1, N: M. The two paths are bidirectionally connected via the relay node device 20.
However, paths other than the 1: 1 connection are branched / merged in the relay node device 20.

In FIG. 12, variable speed terminal device 30
Is a packet assembling circuit 31 and a buffer memory 32 that perform packet transmission, a buffer memory 33 and a packet disassembling circuit 34 that perform packet reception, a packet flow control circuit 35, a buffer amount detection circuit 36, and a speed increase request display transmission circuit. 37 and a speed increase request display receiving circuit 38.

The fixed speed terminal device 40 includes a packet assembling circuit 41 for performing packet transmission and a buffer memory 42.
And a buffer memory 43 for receiving packets, a packet disassembly circuit 44, and a packet flow control circuit 45.

The relay node device 20 includes buffer memories 21 and 22 for bidirectionally relaying packets, route selection circuits 23 and 24, congestion prediction circuits 25 and 26, and speed increase request display erase circuits 27 and 28. Composed.

Next, the operation of the embodiment shown in FIG. 12 will be described. (1) Communication operation when a path is set between the variable speed terminal device 30 and the opposite device Operation of the variable speed terminal device 30 Information input to the variable speed terminal device 30 is packetized by the packet assembling circuit 31. It is digitized and stored in the buffer memory 32. The packet flow control circuit 35 is the buffer memory 3
2 is controlled and the packet is sent to the transmission line 11 at a predetermined packet transfer rate.

On the other hand, the packet arriving from the transmission line 11 is accumulated in the buffer memory 33, and the packet decomposition circuit 34
The original information is restored and output. At this time, the buffer amount detection circuit 36 detects the buffer amount of the buffer memory 33, and if the buffer amount detection circuit 36 detects a buffer amount less than or equal to a predetermined value, the speed increase request display transmission circuit 37 is operated.

If the buffer amount of the buffer memory 33 is equal to or smaller than the predetermined value, the speed increase request display transmission circuit 37 repeatedly sends the packet including the speed increase request display to the transmission line 11 within a predetermined time interval. As the packet including the speed increase request display, a new packet having the speed increase request display is generated, or the buffer memory 32
A speed increase request display area is provided in a part of the packet transmitted from the packet, and the packet for which the speed increase request is displayed is used. Even in the latter case, if there is no packet to be sent even after the lapse of a predetermined time interval, a new packet having a speed increase request indication is generated.

As described above, the variable speed terminal device 30 normally sends out the packet including the speed increase request display at a predetermined time interval or less. On the other hand, if the buffer memory 33 becomes full, the buffer amount detection circuit 36 stops the operation of the speed increase request display transmission circuit 37, and stops the transmission of the packet including the speed increase request display, thereby transmitting the packet. The variable speed terminal device 30 of FIG.

The speed increase request display receiving circuit 38 monitors the arriving packet, and if the packet including the speed increase request display is received within a predetermined time interval, performs the packet transfer speed control performed by the packet flow control circuit 35. maintain. If the packet including the speed increase request indication is not received within the predetermined time interval, it is recognized that there is a possibility of congestion, and the packet flow control circuit 35 is caused to reduce the packet transfer speed at that time. Take control.

Operation of Relay Node Device 20 In the relay node device 20, packets from each terminal device are input to the buffer memories 21 and 22 through the transmission path 11 and accumulated. The congestion prediction circuit 25 detects the total packet transfer rate or packet amount input to the buffer memory 21, or the output packet amount of the buffer memory 21, and activates the speed increase request display erasing circuit 28 when congestion is predicted. To do. Similarly, the congestion prediction circuit 26
Activates the speed increase request display erasing circuit 27. The speed increase request display erasing circuits 27, 28 are connected to the congestion prediction circuits 25, 2
The start control of 6 deletes the packet including the speed increase request display in the reverse direction of the corresponding path or transmission line. As the erasing method, the new packet having only the speed increase request indication is erased, and the packet having the speed increase request indication as a part of the packet is erased only.

In the prediction of congestion in the relay node device 20, it is necessary to predict the round-trip delay time 2d between the farthest variable speed terminal device 30 and the relay node device 20 as a prediction on the safe side, which will be described later. To do. The round-trip delay time 2d includes the control delay time from the detection of the congestion prediction circuit 25 to the reduction of the packet transfer rate by the packet flow control circuit 35, and the delay time in the buffer memory of the relay node device and the like. Is included.

Output from the buffer memories 21 and 22,
The packets that have passed through the speed increase request display / erasure circuits 27 and 28 are output to the transmission line 11 with their output paths selected by the route selection circuits 23 and 24, respectively.

As described above, when congestion is predicted in the relay node device 20 between the variable speed terminal devices 30, the speed increase request display of the path / transmission path related to the congestion is erased, so that the transmission side The variable speed terminal device 30 can be notified of that fact. That is, the variable speed terminal device 30 on the transmission side can recognize that there is a possibility of congestion because the packet including the speed increase request indication will not be received even if it exceeds the predetermined time interval. By reducing the congestion, congestion at the relay node device 20 can be avoided.

(2) Communication operation when a path is set between the fixed speed terminal device 40 and the opposite device The path between the fixed speed terminal devices 40 is a fixed speed service for the transmission path of the path. It can be set when the ratio is equal to or less than the allocated ratio F, and the capacity is secured in the packet multiplexing integrated network. This is used as a path setting allocation standard among all the fixed speed terminal devices 40. Therefore, in the path setting between the new fixed speed terminal devices 40, if there is a portion having a fixed speed service or more with a ratio F or more allocated to any one of the transmission paths of the set path, a path having the ratio F or less is newly added. Either search or set the path. Due to such a limitation, the upper limit occupied by the fixed speed information in the packet multiplex integrated network can be set to the ratio F or less. Then, at least (1-F) can be assigned to communication of variable speed information.

Operation of Fixed Speed Terminal Device 40 The information input to the fixed speed terminal device 40 is packetized by the packet assembling circuit 41 and stored in the buffer memory 42. The packet flow control circuit 45 is the buffer memory 4
2 is controlled and the packet is sent to the transmission line 11 at a predetermined packet transfer rate. The maximum packet transfer rate is the packet transfer rate secured when the path is set.

On the other hand, the packet arriving from the transmission line 11 is accumulated in the buffer memory 43, and the packet decomposition circuit 44
The original information is restored and output.

Operation of Relay Node Device 20 The operation of the relay node device 20 is similar to that of the variable speed terminal device 30, but the operations of the congestion prediction circuits 25 and 26 and the speed increase request display erasing circuits 27 and 28 are fixed. It is irrelevant to the speed terminal 40.

In this way, the fixed speed terminal device 40 is
There is no buffer amount detection circuit 36, speed increase request display transmission circuit 37 and speed increase request display reception circuit 38 provided in the variable speed terminal device 30, and the packet flow control circuit 4 is provided.
Transmission control is performed only by 5. This indicates that the fixed speed terminal device 40 is not affected by the congestion prediction notification in the network. That is, in the communication between the fixed speed terminal devices 40, the packet flow control circuit 4 is irrespective of the state in the network.
Only 5 controls the packet transfer rate. In order to avoid congestion without discarding packets, it is necessary for the relay node device 20 to accurately predict the maximum value of the packet transfer rate. In order to accurately predict the maximum value of the packet transfer rate, the packet transfer rate may be calculated by the acceleration or the upper limit of the acceleration ratio. When acceleration is used, the maximum packet transfer rate increase that depends on the number of terminal devices can be predicted, and when the latter acceleration ratio is used, the maximum packet transfer rate increase that does not depend on the number of terminal devices can be predicted. .

However, in the present embodiment, it is expected that the packet transfer rate in the fixed speed terminal device 40 will be further increased below the speed secured when the fixed speed terminal device 40 sets the path. Therefore, when congestion is predicted in the network, the variable speed terminal device 30 is
If the packet transfer rate is not reduced, including the increase in the packet transfer rate at 0, packet discard occurs. Therefore, it is necessary to clarify the reduction amount of the packet transfer rate.

Next, referring to the graphs shown in FIGS. 13 to 16, the reduction amount of the packet transfer rate required for avoiding congestion will be described.

FIG. 13 is a diagram showing the respective speed relationships of the fixed speed service CBR and the variable speed service VBR used for the prediction of the packet transfer speed in the relay node device 20.

In the figure, the horizontal axis X shows the ratio of fixed speed services (CBR / (CBR + VBR)), the vertical axis S is the throughput, and W is the maximum allowable usage rate without packet discard of each transmission line 11. . It should be noted that in a fixed speed service in which the upper limit occupied by the fixed speed information is F or less, 0 ≦ X
≦ F <1.

Under the condition that the fixed rate terminal device does not reduce the packet transfer rate at the time of congestion prediction, the condition that X = F and the allowable maximum usage rate W does not increase will be obtained. That is, the speed B _c of the fixed speed service CBR is WX, and the variable speed service VBR
The time point (t = 0) at which the speed B _V becomes W (1-X) is predicted, and the condition that the speed does not become higher than that is obtained as follows.

14 and 15 show fixed speed service CB.
It is a figure which shows the time change (acceleration ratio fixed) of each speed of R and variable speed service VBR. In the figure, the horizontal axis t
Is time, and the vertical axis S is throughput. The acceleration ratio due to the passage of 2d time (round-trip delay time between the farthest variable speed terminal device 30 and the relay node device 20) is exp
(2d · Rc) and exp (2d · Rv). Rc is the acceleration ratio coefficient of the CBR terminal, and Rv is the acceleration ratio coefficient of the VBR terminal.

Here, the speed increase δSc of the fixed speed service CBR after the lapse of 2d time is δSc = WX (1-exp (-2d · Rc)), and the variable speed service VBR after the lapse of 2d time is The increase in speed .delta.Sv is .delta.Sv = WX (1-X) (1-exp (-2d.Rv)). Further, since the total rate at t = 0 is the maximum allowable usage rate W, the variable rate terminal device capable of reducing the packet transfer rate is notified at t = -2d. In this variable speed terminal device, since the packet including the speed increase request display does not arrive at the maximum d time delay, the packet transfer speed is multiplied by K (0 <K <1) to reduce the speed, and exp (2d
-Transmit at the acceleration ratio of Rv) and set to W (1-X) K at t = 0. This process is shown in FIG.

As shown in FIG. 16, the reduced amount δSk of the packet transfer rate becomes δSk = WX (1-X) (1-K), and the fixed rate service CBR and the variable rate service VBR after 2d hours have passed. Absorbs the speed increases δSc and δSv. That is, δSk ≧ δSc + δSv.

From the above, the packet transfer rate reduction rate K
Is 0 <K ≦ exp (-2d ・ Rv)-(1-exp (-2d ・ Rc))
If X / (1-X) is satisfied, the total speed at t = 0 will not be the maximum allowable usage rate W. When the acceleration rate of each speed of the fixed speed service CBR and the variable speed service VBR is small (1 >> 2d · Rv> 0, 1 >> 2d · Rc> 0), approximately 0 <K ≦ 1-2d · It becomes Rv-2d.Rc.X / (1-X), and it can be seen that the acceleration ratio of the fixed speed service CBR is expanded by X (1-X) times that of the variable speed service VBR. Therefore, by decreasing the acceleration rate of the fixed speed service CBR, it is possible to reduce the decrease in the maximum allowable usage rate without packet discard on the transmission path.

Further, t = in the relay node device 20
-At 2d, the speed after 2d time is (current speed) x
The maximum speed can be predicted by (the speed increase ratio of the larger of CBR and VBR).

Although the above description shows the case where the speed increase ratio (acceleration ratio) is constant, the same description will be made assuming that the speed increase ratio (acceleration) is constant (Rc for the CBR terminal and Rv for the VBR terminal). be able to. For example, the speed increment of Nc fixed speed terminal devices in 2d time is 2d · Rc · N.
Let c be 2d · Rv · Nv, which is the speed increase of the Nv variable speed terminal devices in 2d time, and L be the decrease of the packet transfer speed when the variable speed terminal device is congested. In this case, if L ≧ 2d · Rc · Nc / Nv + 2d · Rv is satisfied, the increase in packet transfer rate can be suppressed. Further, at the time t = -2d in the relay node device 20, the speed after 2d hours is (current speed) + (CBR and V
The maximum speed can be predicted by the speed increase rate of the larger BR) .multidot.2d.multidot. (Nc + Nv).

By the way, in the above-mentioned embodiment, the congestion avoidance method of the packet multiplex integrated network which provides both the fixed speed service and the variable speed service has been described. It is possible. Further, unlike the method of notifying the terminal device of the congestion prediction signal described in the above embodiment to avoid congestion, in this embodiment, the packet including the speed increase request indication is discarded or the indication is deleted in the network. Therefore, it is possible to avoid congestion without increasing the amount of communication information in the network.

Further, in the present embodiment, the fixed speed terminal device and the variable speed terminal device are separate devices, but one device may be used by switching. For example, the present embodiment can be applied even when a fixed-rate terminal is used up to a certain packet transfer rate and a variable-rate terminal device is used at a higher rate. That is, below a certain packet transfer rate,
Ignoring the operation for the packet including the speed increase request display, the packet is transmitted at the packet transfer speed having the speed increase rate (acceleration) or the speed increase ratio (acceleration ratio) which is equal to or less than the predetermined value, and the variable speed terminal at the speed higher than that It should be operated as.

Further, in the present embodiment, the speed reduction operation of the variable speed terminal device corresponding to the packet including the speed increase request display is an example in which the packet transfer speed of each variable speed terminal device is not fair, but According to the method described in the embodiment, it is possible to impartially use the network of each variable speed terminal device. The packet length may be fixed or variable. However, fixed-length packets are preferable in order to accurately predict congestion at the relay node device 20, but maximum packet lengths (packet lengths at a certain risk rate when the packet length distribution is known) are also used for variable-length packets. It can be used to predict congestion.

(Example of Packet Transfer Rate Prediction) Next, congestion prediction by the congestion prediction circuit 25 of the relay node device or the reception packet terminal device will be described. As described in the above embodiments, according to the present invention, the congestion prediction circuit 25 predicts a future packet transfer rate based on an upper limit value of a predetermined packet transfer rate acceleration or an acceleration ratio, and transfers the packet. When it is predicted that the speed exceeds the allowable speed, a congestion prediction signal is output or the speed increase request display is deleted.

FIG. 17 shows the configuration of an embodiment including a packet rate predictor in the congestion predicting circuit.

The packet rate predictor 53 of the present invention includes a rate measurer 54 for measuring the packet transfer rate after the packets transmitted from a plurality of signal sources 50 corresponding to the transmission packet terminal device are multiplexed by the multiplexing circuit 51. A rate prediction calculation circuit 55 for predicting the packet transfer rate after t hours from the upper limit acceleration α and the current packet transfer rate is provided.

The packet transfer rate prediction by the packet rate predictor 53 will be described. First, for each signal source of the signal source 50 to be measured, the upper limit of the rate of change of packet speed with time (acceleration αi) is defined for each destination address, line number or route identification number. The signal from the signal source is multiplexed by the multiplexing circuit 51 and becomes a multiplexed signal output. This multiplexed signal output is input to the speed measuring device 54. In the speed measuring device 54, the total number of bits P (bit) of all packets per time T (sec) of the current packet flow
From this, the current packet transfer rate V V = P / T (bit / sec) can be measured. Alternatively, the packet interval may be measured as another method of measuring the packet transfer rate. The measured packet transfer rate is input to the rate prediction calculation circuit 55. In addition, the input αi of the rate of change with time of each signal source 50
Then, the prediction time t is input to the speed prediction calculation circuit 55.
This speed prediction calculation circuit 55 performs packet prediction after t time based on these three inputs and outputs it as a speed prediction output.

The prediction calculation in the speed prediction calculation circuit 55 will be described. The speed increase of each signal source 50 after t time is α · t
Since it is limited to the following, the current packet transfer rate will not exceed α · t. Therefore, when the acceleration limit value of the signal source 50 is α _{1 to} α _m , the maximum increase amount is

[0110]

[Equation 8] Is. If this value is added to the current packet transfer rate V _Σ , the maximum rate predicted value V (t) after t hours is obtained.

[0111]

[Equation 9] Packet transfer rate prediction calculation can be done with. Maximum predicted speed value V (t) after this time t and maximum allowable speed V _max that is a threshold value
By comparing with, the congestion after t hours can be predicted.

As described above, according to the present invention, it is possible to guarantee that the maximum predicted velocity is not exceeded by limiting the maximum acceleration of each signal source 50. Since this can be decided deterministically rather than the conventional stochastic prediction, congestion can be reliably avoided.

The maximum velocity prediction by this method is less problematic when the number m of the signal sources 50 is small. However, when the number m is increased, the velocity increase per time becomes large even with a small acceleration. This is because when the congestion control is delayed, especially when the control signal propagation time to the signal source and the delay time d in the above-described embodiment become large, the prediction time increases, and even after the current speed V _now is zero, the prediction time t The device processing speed signal or the transmission line speed may be maximum. This is because the speed prediction by this method depends on the number of signal sources.

Next, a velocity prediction method independent of the number of signal sources will be described with reference to FIG. FIG. 18 shows a configuration of packet transfer rate prediction using a rate-time change rate (acceleration rate: e ^β ).

The multiplexed signal output obtained by multiplexing the signals from the plurality of signal sources 50 is input to the speed predictor 53. The speed predictor 53 includes a speed measurer 56 for each group and a speed prediction calculation circuit 55. The speed prediction calculation circuit 55 performs speed prediction after t hours from the current speed by group measured by the group speed measuring device 56, the speed time change ratio (acceleration ratio: e ^β ) input, and the predicted time t input. .

The group-by-group speed measuring device 56 identifies the input by the destination address, the line number or the route identification number, divides it into n groups, and divides the packet flow groups (G _{1 to} G
_The current packet transfer rates V _{G1 to} V _Gn are measured for each _n ). The current packet transfer rate for each group is output to the rate prediction calculation circuit 55. Each packet transfer rate measurement is the same as the rate measurement of FIG.

The speed prediction calculation circuit 55 performs speed prediction calculation after t hours from the speed time change ratio (acceleration ratio: e ^β ) of each signal source 50 and the prediction time t. If the lower limit of the speed of each signal source is not zero and the speed increase rate after t hours is limited to exp (β) or less per unit time, it will not exceed exp (βt) times the current speed. Therefore, the upper limit of the maximum speed increase is determined. Therefore, for each packet flow group G _{1 to} G _{n in} which the signal source 50 is divided into n groups, the maximum acceleration ratio coefficient β _i (i = 1 to n) in this group and the current packet transfer for each group. From the speeds V _{G1 to} V _Gn , the packet maximum speed V (t) predicted value after t time of the packet flow is

[0118]

[Equation 10] Can be sought by.

To reduce this prediction error, if each packet stream is grouped into n groups, the acceleration ratio exp
It is desirable to divide the level of (β). This is because it is possible to guarantee that the rate of increase will not be exceeded by unifying it to the largest acceleration ratio coefficient β in the group.

Unlike the method shown in FIG. 17, the packet transfer rate prediction method shown in FIG. 18 is suitable for prediction of a multiplexed signal including a large number of signal sources because the prediction formula does not include the number of signal sources. Furthermore, by setting a certain β _i to β _i = 0, it is possible to predict the speed even when a signal of a constant speed is included, and the speed in a packet multiplexing integrated network in which a variable speed signal source and a fixed speed signal source are mixed. It can also be applied to prediction.

The packet transfer rate prediction of FIG. 18 is valid when each signal source 50 constantly generates a signal (when the lower limit of the signal source is not zero). However, the signal source j (j = 1 to 1) that can stop the signal generation (when the lower limit is zero)
s: Let s be the total number of signal sources) and start the signal generation at T ₀ within the prediction time t, the speed of each initial value is VI _j.
Then, the upper limit speed is VI _j · exp (β _j (t−T ₀ )) and the speed increases. If the signal generation start time is random at each signal source, it is impossible to predict. Therefore, when the initial speed VI _i is a very small value, it can be ignored from the speed prediction calculation. If it cannot be ignored, the sum of all initial speeds added to the speed prediction calculation becomes the upper limit value V (t) of the maximum speed prediction, which is expressed by the following equation.

[0122]

[Equation 11] As described above, in the prediction of the packet transfer rate in the embodiment of the present invention, the maximum packet transfer rate after t hours can be predicted, so that future congestion can be accurately predicted in advance, and congestion can be avoided in advance. You can

(Example of Packet Transmission Control) Next, packet transmission control of the transmission packet terminal device 3 will be described. FIG. 19 is a block diagram showing the configuration of the packet transmission control circuit of the packet terminal device. In the configuration of FIG. 5 or 12, the packet flow control circuit 3 of the transmission packet terminal device is shown.
It corresponds to 5. The packet here is a cell that is a fixed length packet.

In FIG. 19, the cell buffer 61 temporarily stores the input cell a, and sends out the stored cell as the output cell c according to the "cell sending present" indicated by the cell sending presence / absence signal b. The cell buffer 61 outputs a cell accumulation signal d when the number of accumulated cells reaches a predetermined number, and outputs an empty signal e when there are no cells to be transmitted. The address generation circuit 62 is a circuit that increments the output address value by 1 in the cell cycle, and when the cell accumulation signal d is given, the read address f that sequentially increases from the initial value 1 is given to the cell transmission interval time table 64. .

Bit information indicating the presence or absence of cell transmission is written in the cell transmission interval time table 64 for each address, and the bit information corresponding to the read address f is output as the cell transmission presence / absence signal b. The interval of bit information indicating the presence of cell transmission corresponds to the cell transmission interval, and each bit information indicating the presence or absence of cell transmission is arranged so that it becomes exponentially shorter. An example of the configuration is shown in FIG. 1 to N are addresses corresponding to the elapsed time for each cell time, and for each address, bit information "1" indicating that the cell is to be transmitted at the elapsed time and the cell to be transmitted at the elapsed time. Bit information "0" indicating that there is no
Is written. Therefore, the cell transmission presence / absence signal b takes a value of "1" or "0" and designates whether or not cells are transmitted at the elapsed time corresponding to each address. As shown here, the interval of "1" (cell transmission interval) becomes exponentially shorter, and a method of calculating the interval will be described later.

The cell speed determiner 63 utilizes the fact that the read address f and the cell transmission interval (cell speed) correspond to each other, and compares with the specified address g to determine whether the cell speed has exceeded the specified value. judge. Further, when it is determined that the cell speed exceeds the specified value, the determination signal h is given to the address generation circuit 62. The address generation circuit 62 also receives a congestion prediction signal i indicating that congestion will be predicted from the network.

The cell transmission control operation will be described based on the above configuration.

When cells are accumulated in the cell buffer 61 and the cell accumulation signal d is given to the address generation circuit 62, the address generation circuit 62 outputs the read address f while incrementing it, and the cell transmission interval time table 64 reads the same. The cell transmission presence / absence signal b corresponding to the address f is sequentially given to the cell buffer 61. As a result, the output cells c are output from the cell buffer 61 at the intervals of the bit information stored in the cell transmission interval time table 64 indicating the presence of cell transmission.
Are transmitted to realize an exponentially shortened cell transmission interval.

When the cell speed corresponding to the cell transmission interval exceeds the specified peak speed, the cell speed determiner 63
Sends the decision signal h to the address generation circuit 62. The address generating circuit 62 returns the read address f to the address value corresponding to the specified cell speed each time the determination signal h is input, and repeats the process of increasing the address value by 1 again. This gives a constant cell velocity (C
BR: Constant bit rate) can be maintained.
In the cell transmission interval time table 64 shown in FIG. 20, assuming that the address N is an address corresponding to the peak speed, and the return address is N-5, the bit information is "1". The peak speed increases. Therefore, in this case, it is better to use, for example, N-4 as the return address as the address that reduces the difference between the average cell speed and the peak speed when the read address f is returned.

When the congestion predicting signal i is input to the address generation circuit 62 from the network, the address value at that time is returned to an address value reduced by a predetermined number or a predetermined ratio, and the address value is increased by 1 again. . This allows the cell speed to be reduced quickly,
It is possible to prevent congestion. When the congestion prediction signal i is input and the address value is reduced by a predetermined ratio, the cell speed is reduced by a constant exponential ratio as shown in the above embodiment, and all packet terminal devices using the network are used. The fair use of can be guaranteed. Further, when the address value is reduced by a predetermined number, the packet terminal device which has transmitted the cell from the beginning is given priority by the method of reducing the cell speed by a certain ratio as shown in the above embodiment.

When the cell transmission presence / absence signal b indicating the presence of cell transmission is given to the cell buffer 61, if the cell in the cell buffer is "0", a new cell is added to the cell buffer from the immediately preceding cell transmission time. The time until the cell is stored is measured, and the address value of the address generation circuit 62 is returned to the cell speed equal to the time width. Alternatively, the address value of the address generation circuit 62 is returned to the initial value.

Next, a method of creating the cell transmission interval time table 64 will be described.

For example, when a constant (initial speed) A of an elapsed time t (sec) from the start of cell transmission and an acceleration ratio coefficient β (1 / sec) are set for a transmission line having a transmission speed V (bit / sec). , The cell speed increases according to the exponential function A · exp (β · t) (bit / sec). The current cell transmission interval is P _i (integer),
If the elapsed time t _i at the start of cell transmission, the cell transmission interval error is E _i (| E _i | <1), and 1 cell time is T ₀ , the next cell transmission interval P _{i + 1} (integer) is V / (A · exp (β · (t _i + Pi · T ₀ ))) + E _i is given by rounding down, rounding up, or rounding off. The packet transmission interval error E _{i + 1} at that time is V / (A · exp (β · (t _i + Pi · T ₀ )) + E _i ).
Given by -P _{i + 1} .

In this way, the cell transmission interval time table 64 can be created by successively obtaining the cell transmission intervals. When calculating the cell transmission interval, the average cell speed is exponentially and smoothly increased over time, especially in the high cell speed region, by correcting the error by taking into account the previous cell transmission interval error. You can

FIG. 21 shows the configuration of another embodiment of the cell transmission control circuit. In the figure, the configuration excluding the address generation circuit 65 and the cell transmission interval time table 66 is shown in FIG.
It is similar to the embodiment of. Cell transmission interval time table 66
When the read address f is given from the address generating circuit 65, the cell sending / receiving signal b corresponding to the read address f is sent to the cell buffer 61, and the return addresses j and k are sent to the address generating circuit 65. To do. The return address j is a return address corresponding to the specified cell rate, and the return address k is a multiplication type or subtraction type return address.

FIG. 22 shows a structural example of the cell transmission interval time table 66. 1 to N are addresses corresponding to the elapsed time for each cell time, and bit information "1" indicating that cell transmission is performed at each elapsed time for each address,
Bit information "0" indicating that cell transmission is not performed is written in the elapsed time. Further, in the cell transmission interval time table 66, return addresses k ₁ to k _N of multiplication type or subtraction type and return addresses j ₁ to j _N corresponding to a prescribed cell rate are written corresponding to each address. . The method of creating the cell transmission interval time table 66 is the same as that of the cell transmission interval time table 64 described above.
However, the multiplication-type return addresses k ₁ to k _N are obtained by rounding down, rounding up, or rounding down after the decimal point of the value obtained by multiplying the address value by the predetermined value K (K <1).

As described above, the address generation circuit 65 stores the return address (multiplication-type return address) k reduced by a predetermined ratio from the current address when the network is used fairly, and at that time. When the congestion prediction signal i is input, the read address f is returned to the return address k and the address increment is restarted. When the network is preferentially used, a return address (subtractive return address) k reduced by a predetermined number from the current address is stored and read when the congestion prediction signal i is input at that time. Address f is its return address k
And the address increment is restarted.

The address generation circuit 65 also stores a return address j corresponding to a specified cell speed, and responds to a judgment signal h output from the cell speed judgment unit 63 when the cell speed exceeds a specified value. Then, the read address f is returned to the return address j and the address increment is restarted. The return address j is set to an address in which the difference between the average cell speed and the peak speed is small.

In this way, the cell transmission interval time table 6
By writing each return address in 6, the address generation circuit 65 can use the given return address as it is without performing necessary calculations. Therefore, although the memory amount of the cell transmission interval time table 66 increases, the multiplier and subtractor in the address generation circuit 65 are not required, and thus high speed operation is possible.

In the embodiment described above, when the congestion prediction signal i is input, the read address f is directly returned to the address reduced by a predetermined number or a predetermined ratio. However, the read at that time is performed. Address f
May be decreased by one every cell time, and the cell speed may be decreased by decrementing to an address decreased by a predetermined number or a predetermined ratio. In this case,
After the target address is reduced, the address increment is restarted.

Further, in the above-described embodiment, when the read-out address f is returned to the address reduced by the predetermined ratio when the congestion prediction signal i is input, the return address is set to the cell transmission interval. Can take any value, that is, the return time position becomes random, so that the probability that cells sent from a plurality of packet terminal devices simultaneously arrive at the congestion relay node device can be reduced.

In the embodiments of FIGS. 19 and 20, the degree of increase in speed is the acceleration ratio exp (β).
In this case, the cell transmission control circuit can be similarly constructed by using the cell interval time table.

[0143]

In the packet network of the present invention, in order to detect the occurrence of congestion in the future, the concept of acceleration or acceleration ratio as a value representing the increase of the packet transfer rate is used for the future packet transfer after a predetermined time. Congestion in the network can be predicted deterministically and accurately based on the current packet transfer rate, and effective packet transmission control can be performed based on that prediction. .
Therefore, (1) there is no packet discard in the packet network,
(2) High throughput network operation, (3) Node buffer memory capacity can be reduced, (4) No limitation on network topology, (5) Integrated network of variable speed and fixed speed communication, (6) ) It is possible to operate as a VBR transmission node at a speed higher than the specified speed of the CBR transmission node.
It is also applicable to N: M communication, and there is an effect that (8) a method of not generating a new packet in the network by the congestion prediction signal notification is possible.

[Brief description of drawings]

FIG. 1 is a diagram for explaining a conventional congestion avoidance principle by stopping transmission at a transmitting node.

FIG. 2 is a diagram illustrating a conventional congestion avoidance principle by window flow control.

FIG. 3 is a diagram showing a form of a packet network to which the present invention is applied.

FIG. 4 is a diagram schematically showing this packet network in a relay node device and a terminal device.

FIG. 5 is a block diagram showing configurations of a variable rate packet terminal device, a fixed rate packet terminal device and a relay node device using the congestion avoidance method of the present invention.

FIG. 6 is a diagram illustrating a method of calculating a packet transfer rate.

FIG. 7: Packet flow rate (1) output from the transmission packet terminal device and packet flow rate input to the relay node device
The figure explaining the time relation with (2).

FIG. 8 is a diagram showing a simulation result of congestion avoidance in the relay node device when k = 1.25.

FIG. 9 is a diagram showing a time relationship between a packet flow rate (1) output from a transmission packet terminal device and a packet flow rate (2) input to a relay node device according to the second embodiment.

FIG. 10 is a diagram showing a simulation result of congestion avoidance in the relay node device.

FIG. 11 is a diagram showing a simulation result of congestion avoidance in the relay node device.

FIG. 12 is a diagram showing configurations of a variable speed terminal device, a fixed speed terminal device, and a relay node device used in the congestion avoidance method of the third embodiment.

FIG. 13 is a diagram showing each speed relationship between a fixed speed service CBR and a variable speed service VBR.

FIG. 14 is a diagram showing a change in speed of a fixed speed service CBR with time (acceleration ratio is constant).

FIG. 15 is a diagram showing a time change of the speed of the variable speed service VBR (acceleration ratio is constant).

FIG. 16 is a diagram showing how the speed of the variable speed service VBR is reduced when it is predicted that the total speed will be the upper limit of the usage rate.

FIG. 17 is a block diagram showing the configuration of a packet rate predictor.

FIG. 18 is a block diagram showing another configuration of the packet rate predictor.

FIG. 19 is a block diagram showing the configuration of a packet transmission control circuit.

FIG. 20 is a diagram showing the contents of a packet transmission time interval table.

FIG. 21 is a block diagram showing another configuration of the packet transmission control circuit.

FIG. 22 is a diagram showing the contents of a packet transmission time interval table.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 transmission line 2, 10, 20 relay node device 3 packet terminal device 11 packet transmission line 21, 22 buffer memory 23, 24 route selection circuit 25, 26 congestion prediction circuit 27, 28 speed increase request display erasure circuit 30 variable speed terminal device 31 packet assembling circuit 32, 33 buffer memory 34 packet disassembling circuit 35 packet flow control circuit 36 buffer amount detecting circuit 37 speed increase request display transmitting circuit 38 speed increase request display receiving circuit 39 congestion prediction signal receiving circuit 40 fixed speed terminal device 41 packet Assembly circuit 42, 43 Buffer memory 44 Packet decomposing circuit 45 Packet flow control circuit 50 Signal source 51 Multiplexing circuit 53 Packet speed predictor 54 Speed measuring instrument 55 Speed predicting calculating circuit 56 Speed measuring instrument by group 61 Cell buffer 62 Address generation times 63 cell rate determiner 64 cell transmission interval table 65 the address generating circuit 66 cell transmission interval table

 ─────────────────────────────────────────────────── ─── Continuation of the front page (31) Priority claim number Japanese Patent Application No. Hei 5-46922 (32) Priority date Hei 5 (1993) March 8 (33) Priority claim country Japan (JP)

Claims (29)

[Claims] 1. A packet network in which packets are transmitted and received between a transmitting node and a receiving node via a relay node, wherein a packet communication speed allowed for the transmitting node is capped. A packet network, wherein an upper limit is set for a value representing a degree of increase in packet communication speed of the transmitting node. 2. The packet network according to claim 1, wherein the value representing the degree of increase of the packet communication speed is a time change rate of the packet communication speed (hereinafter referred to as acceleration). 3. The packet network according to claim 1, wherein the value representing the degree of increase of the packet communication speed is a time change ratio of the packet communication speed (hereinafter referred to as an acceleration ratio). 4. The lower limit or the initial speed is set for the packet communication speed of the transmitting node of the packet network.
The packet network described in any of the above. 5. A relay node or a receiving node in a packet network has an upper limit of a packet communication speed of a passing packet or a packet communication speed of an arriving packet and a value indicating a degree of increase of the packet communication speed. The packet network according to claim 1, further comprising means for predicting a packet communication speed after t hours based on the value. 6. The packet network according to claim 5, wherein the relay node or the receiving node comprises means for predicting that congestion will occur based on the predicted packet communication speed. 7. The packet network according to claim 6, wherein the relay node or the receiving node comprises means for notifying the transmitting node of the congestion prediction when it is predicted that congestion will occur. 8. The present invention is provided in a packet network in which an upper limit is defined for a value indicating a degree of increase in packet communication speed for each destination address or line number or route identification number of a packet transferred on a transmission line. The maximum packet communication speed after t hours based on the detected packet communication speed, the detected packet communication speed, and the upper limit of the value indicating the degree of increase in the specified packet communication speed. A congestion prediction circuit for a packet network, comprising: means for predicting. 9. The upper limit of the acceleration of a specified packet is α
_{When i} (i = 1 to m) and the current total packet rate is V _Σ , the maximum packet communication rate V (t) after t time is 9. The congestion predicting circuit for a packet network according to claim 8, further comprising means for calculating and calculating in accordance with. 10. The upper limit of the specified packet acceleration ratio is exp (β), and there is a lower limit of the packet speed. The transmission path is divided into n groups, and the maximum acceleration ratio exp in the group is divided for each group. From (β _i ) (i = 1 to n) and the current packet communication speed V _{G1 to} V _{Gn of} each group, the maximum packet communication speed V (t) after t hours is calculated by 9. The congestion predicting circuit for a packet network according to claim 8, further comprising means for calculating and calculating in accordance with. 11. An upper limit of a prescribed packet acceleration rate is exp (β), and an initial packet rate is VI.
_j (j = 1 to s), the transmission path is divided into n groups, and the maximum acceleration ratio exp (β
_i ) (i = 1 to n), the current packet communication speed V _{G1 to} V _{Gn of} each group, and the total initial speed of all packet speeds, the maximum packet communication speed V (t) after t hours [Equation 3] 9. The congestion predicting circuit for a packet network according to claim 8, further comprising means for calculating and calculating in accordance with. 12. A transmitting node of a packet network, wherein a packet is transmitted and received to and from a receiving node via a relay node, and an upper limit is set for an allowable packet communication speed of the packet to be transmitted, within a range of the packet upper limit communication speed. , Means for transmitting packets at a communication speed that is a value that indicates a degree of increase in packet communication speed that is less than or equal to a predetermined value, and means for reducing the packet communication speed when receiving a congestion prediction signal indicating that congestion will occur from a relay node or a receiving node A sending node with and. 13. A transmitting node of a packet network, wherein a packet is transmitted and received to and from a receiving node via a relay node and an upper limit is set for an allowable packet communication speed of the packet to be transmitted, within the range of the packet upper limit communication speed. , Means for sending packets at a communication speed that is a value that indicates a degree of increase in packet communication speed that is less than or equal to a predetermined value, and the packet communication speed is reduced even if a congestion prediction signal indicating that congestion will occur from a relay node or a receiving node A sending node with means for not doing. 14. A transmitting node that transmits and receives a packet to and from a receiving node via a relay node, and means for transmitting a packet having a communication speed of a value representing a degree of increase in packet communication speed of a predetermined value or less, and a specified value or less. At the communication speed, the packet communication speed is not reduced even when a congestion prediction signal indicating that congestion is predicted from the relay node or the receiving node, and congestion occurs from the relay node or the receiving node at the communication speed above the specified value. A transmitting node having means for reducing a packet communication speed when receiving a congestion prediction signal indicating that the packet is predicted. 15. A packet sent from a receiving node displaying a packet speed increase request indicating that an increase in receiving speed can be accepted instead of the arrival of a congestion prediction signal when congestion is predicted, to a transmitting node, The transmitting node according to claim 12 or 14, further comprising means for reducing a packet communication speed by not reaching the transmitting node. 16. When a congestion prediction signal is received, at a transfer interval that is k (k> 1) times the packet transfer interval corresponding to the packet communication speed at that time, or at a constant speed or a fixed ratio than the packet communication speed at that time. 16. The transmitting node according to claim 12, 14 or 15, further comprising means for transmitting a packet at a rate at which the packet communication rate is reduced at a constant index ratio. 17. The transmission node according to claim 12, further comprising a packet transmission control circuit that reduces a packet communication speed based on the received congestion prediction signal. 18. A packet transmission interval time table having, as an address for each cell cycle, a time lapse when the packet transmission interval exponentially shortens, and a packet transmission control circuit for performing packet transmission using this table. The transmitting node according to any one of 12 to 17. 19. A packet transmission in which, when the packet communication speed rises to a specified packet communication speed, the time elapse address corresponding to the specified packet communication speed is returned, and from there, the packet is transmitted again using the packet transmission interval time of the table. The transmitting node according to claim 18, further comprising a control circuit. 20. When a congestion prediction signal from the network is received,
20. The transmission node according to claim 18, further comprising a packet transmission control circuit for returning a packet from a time-lapsed address of the table to an address reduced by a fixed number or a fixed ratio and transmitting the packet. 21. A method for avoiding congestion in a packet network, wherein packets are transmitted and received between a transmitting node and a receiving node via a relay node, and an upper limit is provided to a packet communication speed allowed for said transmitting node, The transmitting node sends a packet within a range of an allowable packet communication speed and within a range of an upper limit of a value indicating the degree of increase of the packet communication speed, and the relay node or the receiving node receives a packet of a passing packet. If the packet communication speed after t hours is predicted based on the communication speed or the packet communication speed of the arriving packet and the upper limit of the value indicating the degree of increase of the packet communication speed, and congestion is predicted, Notifying the transmitting node that congestion is predicted, and when the transmitting node receives the notification that congestion is predicted, at that point in time Congestion avoidance method for packet network, characterized by reducing the packet transmission rate. 22. A packet is transmitted and received between a transmitting node and a receiving node via a relay node, an upper limit is set for the packet communication speed of all the transmitting nodes, and the transmitting node predicts congestion. Upon receiving the notification,
A transmitting node that reduces the packet communication speed, a transmitting node that does not reduce the packet communication speed, and a transmitting node that reduces the packet communication speed during transmission at or above the specified speed but does not reduce the packet communication speed during transmission at or below the specified speed. An integrated network that includes a packet transmission speed reduction amount that reduces the packet communication speed when receiving communication indicating that congestion is predicted, and reduces the amount of speed increase prediction including the transmission node that does not decrease speed. Packet network congestion avoidance method. 23. The congestion avoidance method for a packet network according to claim 21, wherein the value representing the degree of increase in packet communication speed is acceleration. 24. The congestion avoidance method for a packet network according to claim 21, wherein the value representing the degree of increase in packet communication speed is an acceleration ratio. 25. The congestion avoidance method for a packet network according to claim 21, wherein the packet communication speed is reduced by a constant value when the notification that congestion is predicted is received. 26. The packet communication speed is reduced by a fixed ratio when the notification that congestion is predicted is received.
27. A packet network congestion avoidance method according to any one of 1 to 24. 27. Upon receiving a notification that congestion will occur,
The packet communication speed is reduced by a constant index ratio.
25. The packet network congestion avoiding method according to any one of 1 to 24. 28. The packet network congestion avoidance method according to claim 21, wherein the notification that congestion occurs is performed by not transmitting a packet communication speed increase request signal or an increase request indication to the transmitting node. . 29. The packet network congestion avoidance method according to claim 23, wherein the packet communication speed of the transmitting node is set to a lower limit or an initial speed.