JP3079066B2 - ATM communication network - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to an ATM (Asynchronou).
s Transfer Mode) Used for communication. The present invention relates to a technique for variably setting the bandwidth of a VC route according to traffic.

[0002]

2. Description of the Related Art In an ATM communication network, a VP link is set on a physical transmission line provided between exchanges in the network.
Further, a VC route is set for this VP link. A plurality of VC routes can be set in one VP link. Further, the bandwidth of the VC route can be set arbitrarily within a range in which the sum does not exceed the maximum capacity of the VP link accommodating the VC route.

In recent ATM communication networks, in order to make effective use of the bandwidth of the VP link, the bandwidth of the VC route is varied according to the traffic at that time.

In the bandwidth control of a VC route in a conventional ATM communication network, the number of cells waiting for output in a buffer in an exchange is observed, and the observation result is notified to an exchange accommodating the transmission end, so that the transmission end is monitored. The switch that accommodates the cell performs the output rate control of the cell by grasping the congestion state.

[0005] Regarding the control of the output rate of such a cell, only past information can be obtained by notifying observation results due to propagation delay, and information is given only at an appropriate notification interval. It is necessary to increase the buffer amount or to reduce the response speed to the bandwidth change request from the user as the bandwidth increases.

For example, in the ABR system, RM (Resource Ma
nagement) Output rate control is performed using a control cell called a cell. The upstream switch controlling the output rate uses the RM cell to check the congestion situation in the network.
In other words, the upstream exchange controlling the output rate inserts the RM cell every fixed number of cells and transmits it to the receiving end.

A downstream exchange whose output rate cannot be controlled writes information on the presence or absence of congestion and an acceptable input rate in the RM cell. The RM cell is looped back at the destination exchange or the receiving end and returns to the exchange controlling the output rate upstream.

[0008] Therefore, an exchange which can control the output rate by reading the information on the returned RM cell,
The output rate can be controlled according to the congestion situation in the network. However, since there is generally a propagation delay between the switch performing the observation and the switch controlling the output rate, a buffer for absorbing congestion is required until the notified information is reflected in the actual output rate reduction. It is necessary to limit the bandwidth increase of each connection so that congestion does not progress rapidly or in a short time. In particular, WAN (Wide
In networks with long delays, such as Area Networks, this problem is significant, making service difficult.

In the ABR system, the number of output waiting cells in a buffer in an exchange is observed, and a congestion state is notified to the originating exchange. This conventional example is shown in FIG. FIG. 15 is a diagram showing the concept of an ATM communication network based on the ABR system. In FIG. 15, when the queue lengths of the exchanges 3 and 4 that are observing exceed a predetermined threshold value, the exchange 3 notifies the originating exchange 1 of congestion, and can control the output rate. 1 decreases the rate by a predetermined amount when congestion information arrives, and conversely, when it is notified that it is not congestion,
The output rate is controlled by increasing the rate by a predetermined amount.

FIG. 16 and FIG. 17 show this processing flow. FIG. 16 is a diagram illustrating a processing flow of the exchange 1 that performs output rate control. FIG. 17 is a diagram showing a processing flow of the exchanges 3 and 4 that perform observation. As shown in FIG.
The exchanges 3 and 4 performing the observation check the number of cells in the buffer (S14), and if the number of cells exceeds the threshold (S15), notify the congestion state using the RM cell (S16). . If the number of cells does not exceed the threshold (S15), the normal state is notified using the RM cell (S17). As shown in FIG. 16, the exchange 1 that performs output rate control receives the notification from the exchanges 3 and 4 as control information (S10), and if the notification is a notification of congestion (S11), reduces the output rate. If the notification is not a notification of congestion (S11), the output rate is increased (S12).

In this technique, a congestion signal for lowering the output rate is transmitted after congestion occurs. In addition, since the increase rate of the output rate is set to be constant, the output rate increases by the same amount until immediately before the congestion occurs. For these reasons, the reduction rate of the output rate must be set large. This means that, in other words, since it is based on only the information on the presence or absence of congestion, control according to the degree of congestion is not performed.

[0012]

In such a conventional output rate control, since there is a distance between the exchange for controlling the output rate and the exchange for performing the observation, even if the congestion notification is performed, the output rate is actually changed. There is a delay time until control is started, during which time output rate control to recover from a congested state cannot be expected, and cell traffic for overload during that time is absorbed by the buffer, so an extremely large buffer Need quantity.

[0013] As the delay time between the exchange controlling the output rate and the exchange performing the observation increases, the system is exposed to the congestion state for a long time, so that a larger amount of buffer is required. In order to prevent this, in networks with large transmission delays, the output rate should be set so that the rate of increase is reduced and the rate of decrease is increased, thereby causing rapid congestion and recovering from congestion in a short time. Can be considered.

However, this setting may lower the transmission rate because the output rate assigned to the connection slowly rises and drops immediately.

A method of calculating an acceptable input rate in each exchange and notifying the exchange controlling the value is an ATM.
It is shown in Forum 95-0013. However, in this conventional method, it is necessary to calculate an input rate that can be accepted by all the exchanges in the network, and to notify the controlling exchange of the minimum value among them.

Therefore, it is difficult to calculate an acceptable input rate because it is necessary to perform complicated calculations at high speed. Also, in this conventional method, when the delay time is long, the increase rate of the output rate is reduced and the decrease rate of the output rate is increased. Further, the acceptable input rate is given to the exchange controlling the output rate of the calling side via the RM cell, so that until the next RM cell arrives, the rate specified from the nearest old RM cell is used. The output at is forced by the exchange to control the output rate. Therefore, it is assumed that there is no change in the network congestion state until the next RM cell arrives, and control is performed as it is.

In an actual communication network, there are a switch near the switch controlling the output rate and a switch far from the switch. Changing the control parameters in accordance with the distance and the delay time is very troublesome, and the connection is difficult. It is difficult to measure the delay each time. Furthermore, many VC routes are multiplexed in one switch, and changing the control parameters for each of them is not preferable because the complexity of setting the control increases.

It is an object of the present invention to provide an ATM communication network which has been made in such a background and which enables an observing exchange to generate output rate control information without requiring complicated calculations. With the goal. SUMMARY OF THE INVENTION It is an object of the present invention to provide an ATM communication network which can avoid congestion without being affected by a delay time in an exchange controlling an output rate. SUMMARY OF THE INVENTION An object of the present invention is to provide an ATM communication network capable of rapidly increasing an output rate if there is an available band. SUMMARY OF THE INVENTION It is an object of the present invention to provide an ATM communication network capable of performing safe output rate control while leaving a desired band.

[0019]

According to the present invention, an input load factor is observed in an exchange located downstream of a VC route, and the observation result is notified to an upstream exchange capable of controlling an output rate.
Based on the information, the switch capable of controlling the output rate calculates the output rate for the purpose of outputting cells at the highest possible output rate while avoiding congestion in the downstream switch, and calculates the calculated output. It is characterized by outputting cells at a rate.

The difference from the prior art is that the switch performing the observation observes the input load factor and notifies the exchange that controls the output rate of the observation result. Another difference is that the exchange that controls the output rate receives the notification of the input load rate, calculates the output rate from the received input load rate, and controls the output rate to be equal to or less than the calculated rate.

Further, in the conventional technique, when a switch which performs output rate control is notified from a switch which performs observation, a switch between the receivers receives information on an input load factor from downstream, and Is different from the observed input load factor in that the larger value is reported upstream.

The applicant of the present application has filed a prior application (Japanese Patent Application No. Hei 8-2255).
No. 96, which was not disclosed at the time of filing the present application), the remaining bandwidth ratio is used as notification information, but the present invention differs in that the input load ratio is used. Also, the procedure for obtaining the rate increase / decrease ratio from the notification information is different. Another difference is that an extrapolation operation is performed based on the notified information to perform an input load factor calculation for the purpose of estimating the current input load factor in the VC route on the safe side. Also, the output rate control flow is different.

That is, the present invention comprises an exchange accommodating a transmitting end and a receiving end, a physical transmission line connecting the exchanges, and a relay exchange inserted through the physical transmission line. V between which a VC route can be set
An ATM communication network in which P links are respectively set. According to a feature of the present invention, the exchange has a means for calculating and recording its own input load factor (input rate / VP link capacity), and the input load factor is transmitted to the exchange accommodating the transmitting end. Means for sending as input information, and when there is input load factor information arriving from another switch, comprising means for comparing the input load factor recorded by itself with the input load factor included in the incoming input load factor information, The sending means includes means for sending the larger input load factor as the latest input load factor information according to the comparison result of the comparing means.

It is preferable that the exchange accommodating the transmitting end has means for variably setting a VC route band set by its own exchange according to the comparison result of the comparing means.

As a result, the observing exchange can generate the input load factor information as the output rate control information without requiring a complicated calculation.

The variable setting means estimates the current input load factor until the next input load factor information arrives in accordance with the input load factor included in the previous input load factor information that has arrived. It is desirable to include

The estimating means includes an estimated value of input load factor Ψpred, where γ is an output rate increase / decrease ratio, P1 and P2 are parameters for determining characteristics of the output rate increase / decrease ratio γ, and R is an output rate. Γ = P1 * Ψpred + P2 R = R * γΨpred = Ψpred * γ.

Thus, the exchange which controls the output rate even while the input load factor information does not arrive can control the bandwidth of the VC route at the optimum output rate.

It is desirable to have means for updating the estimated value in accordance with the estimated value and the input load factor included in the newly arrived input load factor information.

The means for performing the update includes setting the input load factor included in the newly arrived input load factor information to Ψnotified, setting the initial value of the variable π to “1.0”, and setting the value of the variable π to the value of the previous variable π. It is preferable to include means for sequentially calculating the values by multiplying the values by γ, and calculating the updated estimated value Update @ prep as Update @ prep = min ($ pred, $ notified * π). Thereby, updating of the estimated value can be regarded as updating on the safe side.

[0031]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The structure of an embodiment of the present invention will be described with reference to FIGS. FIG. 1 is a diagram showing the overall configuration of an embodiment of the present invention. FIGS. 2 and 3 are block diagrams of the exchange of the embodiment of the present invention.

The present invention relates to an ATM communication network, wherein
Exchange 1 accommodating -1, 5-2 and receiving end 6-1
1, 1-2, 2-2 and the exchanges 1-1, 1-2, 2
-2, a physical transmission line (not shown) for connecting the two, and a relay exchange 2-1 interposed in the physical transmission line. VC between -2
VP link 3-0 in which routes 4-1 and 4-2 can be set,
ATM communication networks 3-1, 3-2, 3-3 and 3-4 are respectively set.

Here, a feature of the present invention is that the exchanges 1-1, 1-2, 2-1 and 2-2 calculate and record their own input load factors (input rate / VP link capacity). Arrival cell number counter 31, arrival rate calculator 32, input load factor calculator 33, and RM as means for transmitting the input load factor as input load factor information to exchange 1-1 accommodating the transmitting end. When there is input load factor information arriving from another switch, the cell transmitting unit 35 and a comparison operation unit as a means for comparing the input load factor recorded by itself with the input load factor included in the incoming input load factor information 34, and the RM cell transmitting unit 35 transmits the larger one of the input load factors as the latest input load factor information according to the comparison result of the comparison operation unit 34.

The exchanges 1-1 and 1-2 accommodating the transmission ends 5-1 and 5-2, according to the comparison result of the comparison operation unit 34, set the VC routes set by their own exchanges 1-1 and 1-2. An output rate calculator 38 and an output rate controller 37 are provided as means for variably setting the bands 4-1 and 4-2.

The output rate calculator 38 estimates the current input load factor until the next input load factor information arrives according to the input load factor included in the previous input load factor information that has arrived.

This estimation is based on the assumption that the output rate increase / decrease ratio is γ,
Assuming that the parameters for determining the characteristics of the output rate increase / decrease ratio γ are P1 and P2 and the output rate is R, the estimated value 率 pred of the input load factor is expressed as γ = P1 * １pred + P2 R = R * γ Ψpred = Ψpred * γ Is calculated as

The output rate calculating section 38 updates the estimated value according to the estimated value Ψpred and the input load factor included in the newly arrived input load factor information.

In this update, the input load factor included in the newly arrived input load factor information is set to Ψnotified, the initial value of the variable π is set to “1.0”, and the value of the variable π is set to γ to the value of the previous variable π. Multiplied by, and the updated estimated value Update
Calculate Ψprep as UpdateΨprep = min (Ψpred, Ψnotified * π).

[0039]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiments of the present invention will be described. In FIG. 1, 1-1 and 1-2 are exchanges for controlling the output rate, 2-1 and 2-2 are exchanges for controlling the input load factor without controlling the output rate, 3-0, 3-1 and 3-2. Reference numerals -2, 3-3, and 3-4 denote VP links set between the exchanges 1-1, 1-2, 2-1 and 2-2, and 4-1 and 4-2 denote VC routes.

The exchanges 1-1, 1-2, 2-1 and 2-2 observe the input rates applied to the VP links 3-1, 3-2, 3-3 and 3-4, respectively. The input load factor in the link is determined by (input rate / VP link capacity).

The exchange 2-2 notifies the exchange 2-1 of the input load factor for the VP link 3-4. Exchange 2-2
Notifies the exchange load factors of the VP link 3-4 to the exchanges 1-1 and 1-2. Further, the exchange 2-1 extracts the input load factor information from the exchange 2-2, and determines the larger one of the input load factor of the VP link 3-3 and the value of the input load factor notified from the exchange 2-2. Exchanges 1-1 and 1
-2 is notified.

This is because the VC route 4 is transmitted from the receiving end 6-1.
RM cell for control is transmitted every 1 and 4-2,
Writing the input load factor on the M cell, rewriting the value of the input load factor by the above procedure and relaying the data to the exchanges 1-1 and 1-2 each time the signal passes through each of the exchanges 2-2 and 2-1. Is realized by:

FIG. 2 shows a block diagram of the exchanges 2-1 and 2-2. In the exchange illustrated in FIG. 2, consider measurement and notification of an input load factor related to a cell flow flowing from left to right in a certain VC. In FIG. 2, the route 11 is the input side of the VP link into which the cell flow of interest flows, and the route 12
Represents a side opposite to the path 11 when two-way communication is considered using the same VP link as the path 11. Similarly, the path 21 is the output side of the VP link from which the cell flow flows, and the path 22 is the path 2
Path 21 when two-way communication is considered on the same VP link as 1
Represents the opposite direction.

In FIG. 2, the number of arriving cells for each fixed observation time is calculated by an arriving cell counter 31, and the value is sent to the arrival rate calculating unit 32. In the arrival rate calculation unit 32, the arrival rate (Rate) is calculated as: Rate = α * number of arrival cells / observation time + (1−α) Ra
te is calculated using the following equation. Then, the input load factor 到 着 is obtained from the maximum value of the arrival rate observed in the window delimited at regular intervals as Ψ = maximum value of Rate / output link capacity. The obtained input load factor Ψ is sent to the input load factor calculation unit 33. The input load factor calculator 33 sends the input load factor Ψ to the comparison calculator 34.

The RM cell extractor 36 extracts the RM cell arriving from the path 22, extracts the input load factor information written on the RM cell, and sends the input load factor value to the comparison calculator 34. The RM cell transmitting unit 35 compares the input load factor calculated by the input load factor calculating unit 33 with the input load factor included in the input load factor information of the RM cell, and passes the larger value to the RM cell transmitting unit 35. . The RM cell sending unit 35 writes the received value into the RM cell and sends it out.

In FIG. 1, exchange 1-1 is exchange 2-
The larger of the input load factor notified from No. 1 and the input load factor of the VP link 3-1 is held as the total input load factor of the VC route 4-1. Thereby, the exchange 1-
1 can obtain information on the maximum input load factor on the VC route 4-1 through which traffic flows. The exchange 1-2 similarly holds the input load factor of the VC route 4-2.

FIG. 3 shows a block diagram of the exchanges 1-1 and 1-2. In the exchange shown in FIG. 3, measurement and notification of an input load factor related to a cell flow flowing from left to right of a certain VC route will be considered. Path 41 is the input side of the VP link into which the cell flow of interest flows, and path 42 is the path 41
When the two-way communication is considered with the same VP, it represents the direction opposite to the route 41. Similarly, the route 51 is the output side of the VP link from which the cell flow flows, and the route 52 is the same V as the route 51.
When the two-way communication is considered by the P link, it represents the opposite side from the path 51.

It is assumed that the exchange shown in FIG. 3 is connected to the exchange 2-1 by a VP link. Measurement of input load factor
This can be performed in the same manner as described with reference to FIG.
The operations of the RM cell extraction unit 36 and the comparison operation unit 34 are the same as those in FIG. The input load ratio output from the comparison operation unit 34 represents the overall input load ratio of the VC route, and the value is sent to the output rate calculation unit 38. The output rate calculator 38 calculates the output rate, and the output rate controller 37 receives the value and sets the output rate to a value lower than the value.

The operation of the output rate calculator 38 will be described with reference to FIGS. 4 and 5 are flowcharts showing the operation of the output rate calculation section 38. The output rate calculator 38 calculates the output rate from the input load factor of the VC route at an appropriate cycle. Output rate calculator 38
Is performed by a combination of the two processing flows shown in FIG. 4 and FIG.

In the processing flow shown in FIG. 4, every time a cell arrives (S0), it is checked whether or not the timer has exceeded a set time (S1). The following calculation for estimating the load factor is performed.
Assuming that the output rate increase / decrease ratio is γ, the parameters for determining the characteristics of the output rate increase / decrease ratio γ are P1 and P2, and the output rate is R, the input load factor estimated value Ψpred is given by γ = P1ΨΔpred + P2 R = R · γ と し て pred = Ψpred · γ π = π · γ The output rate dynamically changes by multiplying the value of the output rate increase / decrease ratio γ. Also, the variable π is the increase / decrease ratio γ of the output rate from the previous notification to the present.
Is a value for collectively expressing.

The initial value of the variable π is set to “1.0”.
This value is used in the processing flow shown in FIG. 5 when the next notification is received. Then, after those calculations, the timer is reset (S2), and the arrival of the next cell is waited.
When the cell arrives, if the set time of the timer has not been exceeded, nothing is waited for the arrival of the next cell.

In the processing flow of FIG. 5, when a cell arrives (S3), it is checked whether or not the cell holds new input load factor information (S4). An operation of (Ψpred, ＊ notified * π) π = 1.0 is performed (S5). Here, Ψnotified represents the value of the input load factor newly notified. If the arriving cell does not hold the new input load factor information, it waits for the next cell to arrive.

Next, the operation principle of the processing flow of FIGS. 4 and 5 will be described. The purpose of this processing flow is to control the processing flow without increasing the required buffer amount and lowering the transfer characteristics by predicting congestion discretely based on the notification information of the usage rate.

First, in order to explain the operating principle of the processing flow of FIGS. 4 and 5, consider the control system of FIG. FIG. 6 is a diagram showing a control system for explaining a control procedure of the output rate. In FIG. 6, control nodes (# 1) to (#n)
Pay attention to the traffic flowing from to the observation node.

It is assumed that the traffic entering the observer node is multiplexed towards its single output link. At this time, the observation node observes the total traffic T flowing in, and calculates the ratio of the total traffic T to the capacity C of the output link by the input load ratio Ψ
Asking. That is, the input load factor at the observation node in FIG. 6 can be defined by Ψ = T / C (2.1). The observation node calculates the input load factor Ψ
To the upstream control nodes (# 1) to (#n). Here, for the sake of simplicity, it is assumed that the notification is made at a constant period.
The control node has a function of dynamically changing the output rate based on the notified information. The control node controls the output rate to achieve high link utilization and avoid congestion.

To explain the principle of operation, it is assumed that the utilization at the observation node is always known by the control node. Also, consider an ideal system with no delay between the observation node and the control node. In this system, consider that the control node dynamically controls the output rate by repeatedly calculating the following two equations.

R = R * γ (2.2) γ (Ψ) = P1 * Ψ + P2 P1 ≦ 0, P2> 1.0 (2.3) where R is the output rate from the control node, and γ is the output Output rate increase / decrease ratios for controlling the rate, P1 and P2, respectively represent parameters for determining characteristics of the rate increase / decrease ratio.

P1 takes a non-positive value and P2 takes "1.
By taking a value larger than 0 ", the output rate increase / decrease ratio γ becomes a non-increasing function of the input load ratio Ψ.
As the input load ratio increases, the output rate increase / decrease ratio γ decreases. The input load factor Ψ is the output rate R of all control nodes.
Is represented by the sum of Therefore, when the input load factor Ψ is “1.
When the value is near “0”, the output rate increase / decrease ratio γ is set to be “1.0”, and by repeating the calculations of Expressions (2.2) and (2.3), greedy traffic is obtained. , The value of the input load factor 収束 can converge to 1.0.

Actually, various initial values are set, and the equation (2.
FIG. 7 shows an example in which the calculations in 2) and (2.3) are performed.
FIG. 7 shows the input load factor and equations (2.2) and (2.3).
It is a figure which shows the relationship of the number of calculations of. The horizontal axis indicates the number of repetitions of the calculation, and the vertical axis indicates the input load factor. P1 was set to "-0.5", and P2 was set to "1.5". here,
Greedy traffic means a traffic source that has the property of always trying to match its average utilization with the declared load. In FIG. 7, the desired average utilization rate of the traffic source is set to “2.0”.

FIG. 7 shows that the input load factor Ψ converges to “1.0” by repeating the calculation even when the initial value of the input load factor Ψ is variously changed. Since the output rate increase / decrease ratio γ is a function that defines the maximum output rate increase / decrease ratio, traffic does not increase beyond this ratio. Therefore, it is possible to perform control at a high usage rate and to avoid congestion.

Further, under the assumption that the control node always grasps the fluctuation of the input load factor 、 and the delay between the nodes is “0”, even if the control node temporarily becomes non-greedy,
Immediately, the current value of the input load factor Ψ is obtained, and control using this value as an initial value starts, so that the operation of bringing the usage rate closer to “1.0” is performed again.

Since the equation (2.3) is a decreasing function of the input load ratio Ψ, the smaller the input load ratio Ψ, the larger the value of the output rate increase / decrease ratio γ. That is, when the load is low, the increase / decrease ratio γ of the output rate becomes large, indicating that the band can be accessed at high speed. Conversely, when the load is high, there is a strong tendency to increase the output rate slowly. Also, as the degree of overload increases,
Since the increase / decrease ratio γ of the output rate is small, the tendency is that the output rate is reduced as soon as severe congestion occurs.

Under the assumption that the control node can always grasp the input load factor of the VC route without delay, the equation (2.
The control was possible only by the repetitive calculation of 2) and (2.3). Here, as a more practical assumption, a case is considered in which input load factor information is carried by control cells transmitted at appropriate intervals. Note that the notification interval is shorter than the output rate update interval in the control node.

At some point, if the current input load factor 観 測 at the observation node is clear, the maximum input load factor after the control node updates the output rate once is: Ψpred = Ψ · γ (Ψ) ( It is easily inferred from 2.4).

Since the output rate increase / decrease ratio γ given by the equation (2.3) is a function of the input load rate Ψ, the input load rate of the equation (2.4) is estimated instead of the input load rate い な い which is not yet known. If the value Ψpred is used, the increase / decrease ratio γ of the output rate with respect to the maximum input load ratio is obtained. Thereafter, by repeating this calculation, extrapolation of the worst traffic (Extrapolati
on) It can be obtained by calculation. Therefore, even with this change, the behavior of the usage rate in the case of greedy traffic matches the curve in FIG. In addition, even when it is not greedy, the estimated value of the input load factor Ψ
By replacing pred with the input load factor Ψ obtained from the new notification, the control information can be corrected again.

By definition, if the traffic is not greedy, there is a point that Ψ <Ψpred. At that time, since γ (Ψ) <γ (Ψpred) holds from the equation (2.3), the value calculated by this estimation includes only the error on the safe side.

When the delay between the control node and the observation node is considered, the notified input load factor information has already deviated from the current state. Therefore, using the information as it is reduces the control efficiency.

Therefore, this problem can be reduced by the control node executing the following control algorithm. This control algorithm is realized by a combination of the two processing flows shown in FIGS.

However, in the initial state, the input load factor Ψ
Is given. The processing flow shown in FIG. 4 estimates the input load factor and updates the output rate based on the above-described extrapolation calculation. The processing flow shown in FIG. 4 is repeatedly executed until the next input load factor information is received. In this processing flow, the variable π is a value for collectively expressing the increase / decrease ratio γ of the output rate from the reception of the previous input load factor information to the present. The initial value of the variable π is set to “1.0”. This value is used in the processing flow shown in FIG. 5 for the next reception of the input load factor information.

The operation of the processing flow shown in FIG. 5 will be described with reference to FIGS. 8 and 9 are diagrams for explaining the operation of the processing flow shown in FIG. 5, in which the horizontal axis indicates the propagation delay time, and the vertical axis indicates the input load factor Ψ and the estimated value of the input load factor Ψpred. . The observation node is an input load factor (Ψnotified) at a certain observation time point (tobs).
To the control node. However, the control node can receive the information later by the propagation delay.

In FIGS. 8 and 9, the time when the control node receives the notification is represented by t _N−1 and t _N (where t _N−1 <
t _N ). In the processing flow shown in FIG. 5, when the notification is received, the estimated value Ψpred of the input load factor is calculated by Ψpred = min (Ψpred _, Ψnotified * π) (2.5). According to equation (2.5), the estimated value Ψpred input load factor at time t _N in FIG. 8 is at higher than Ψnotified * π employs Ψnotified * π.

As shown in FIG. 8, the estimated value Ψpred of the input load factor is on the safe side because the maximum traffic is assumed. However, since the actual input load factor か ら is at most π times from the time of notification, as shown in FIG. 8, を notified * π is replaced with the new input load factor estimated value Ψpred.
Even if it is adopted, it will be an update on the safe side. Similarly, the estimated value Ψpred of the input load factor at the time t _N in FIG.
When it is smaller than ed * π, the estimated value Ψpred of the input load factor calculated by the processing flow shown in FIG. 4 is continuously used. Also in this case, the estimated value Ψpred of the input load factor is a value that predicts the worst traffic from the discussion so far, and thus the estimated value Ψprd of the input load factor according to equation (2.5).
It is safe to adopt as ed.

As described above, the control algorithm shown in FIG. 4 and FIG. 5 is based on the fact that the control node dynamically changes the rate based on the information on the input load ratio notified from the observation node. Notifications are delayed,
Even when the information can be obtained only at appropriate intervals, safe output rate control can be performed.

As described above, in the exchange performing the observation, the input load factor is observed, the value is notified to the exchange controlling the output rate on the upstream side, and the exchange controlling the output rate notified receives the value. , The input load factor of the current VC route is estimated based on the value, and the increase or decrease of the output rate is controlled based on the value. Therefore, high throughput can be achieved without causing congestion. You can also explicitly measure the delay between exchanges,
A high throughput can be achieved without congestion without grasping.

FIG. 10 shows a comparison of the network throughput between the ABR system and the system of the present invention. The horizontal axis represents the input load, and the vertical axis represents the network throughput.
Input load is normalized by link capacity, and network throughput is normalized by input load. FIG. 10 shows a case where the burst length is 32 cells and a case where the burst length is 128 cells, respectively, for the system of the present invention and the ABR system. Here, the burst length refers to the number of cells when a plurality of cells are collectively transferred in a short time.

At a low load, the throughput is almost the same, but at a high load, the present invention shows a higher throughput.

Further, since the value of the input load indicates the risk of congestion, the rate of increase is automatically set to a small value in a state of a high input load, and is automatically set in a state of a low input load. Because the rate of increase is set to be large, the output rate rises quickly in an environment with a low input load,
Conversely, under an environment with a high input load, it is possible to control the output rate safely without congestion. As a result, an effect of suppressing an increase in the cell buffer queue length under a high load can be provided.

FIG. 11 shows the input load and the maximum cell buffer capacity.
This shows the relationship between queue lengths. The horizontal axis indicates the input load, and the vertical axis indicates the maximum cell buffer queue length. Input load is normalized by link capacity. In FIG. 11, the burst length is 3 for each of the system of the present invention and the ABR system.
The case of two cells is shown.

In the ABR system, the queue length of the cell buffer increases as the input load increases, but it can be seen that the effect of the input load is small in the system of the present invention.

FIG. 12 shows the relationship between the input load and the burst transfer rate. The horizontal axis indicates the input load, and the vertical axis indicates the burst transfer rate. The input load is normalized by the link capacity, and the burst transfer rate is normalized by the link capacity. The burst transfer rate is a value indicating the output rate given to the arriving burst. FIG. 12 shows a case where the burst length is 32 cells and a case where the burst length is 128
The case of a cell will be described. FIG. 13 shows how the transfer time is reduced by the burst transfer rate. The horizontal axis represents time, and the vertical axis represents burst transfer rate. When the burst transfer rate is high, the burst transfer is completed in a short time as shown in FIG. 13, so that the quality of service for the user is improved.

FIG. 12 shows that the system of the present invention shows a high burst transfer rate particularly under an environment of a low input load, and that the output rate can be quickly increased.

In addition, since the node that performs the observation only needs to calculate the input load factor, the calculation load on the network, especially the transit exchange, can be reduced.

Further, in the method of the present invention, even when the delay time information is not explicitly used, the estimation of the input load factor is corrected and controlled on the safe side. It can respond to various network scales without adjustment of.

FIG. 14 shows the relationship between the propagation delay time and the maximum cell buffer queue length. The horizontal axis indicates the propagation delay time, and the vertical axis indicates the maximum cell buffer queue length. FIG.
4 shows a case where the burst length is 32 cells and a case where the burst length is 128 cells, respectively, for the system of the present invention and the ABR system.

Since the ABR system does not include a control in consideration of the propagation delay, the buffer queue length increases unless the parameter adjustment is changed according to the delay. On the other hand, in the method of the present invention, the output rate can be controlled without changing the parameters and without being affected by the transmission delay.

[0086]

As described above, according to the present invention,
The observing switch can generate output rate control information without requiring complicated calculations. In addition, the switching device that controls the output rate can avoid congestion without being affected by the delay time, and can increase the output rate at high speed if there is an available bandwidth. Further, it is possible to perform safe output rate control while leaving a desired band.

[Brief description of the drawings]

FIG. 1 is a diagram showing an overall configuration of an embodiment of the present invention.

FIG. 2 is a block diagram of the exchange according to the embodiment of the present invention.

FIG. 3 is a block diagram of the exchange according to the embodiment of the present invention.

FIG. 4 is a flowchart showing the operation of an output rate calculation unit.

FIG. 5 is a flowchart showing the operation of an output rate calculation unit.

FIG. 6 is a diagram showing a control system for explaining a control procedure of an output rate.

FIG. 7 is a diagram showing a relationship between an input load factor and the number of calculations.

FIG. 8 is a diagram for explaining the operation of the processing flow.

FIG. 9 is a diagram for explaining the operation of the processing flow.

FIG. 10 is a diagram showing a comparison of network throughput between the ABR method and the method of the present invention.

FIG. 11 is a diagram showing a relationship between an input load and a maximum cell buffer queue length.

FIG. 12 is a diagram showing a relationship between an input load and a burst transfer rate.

FIG. 13 is a diagram illustrating a manner of shortening a transfer time by a burst transfer rate.

FIG. 14 is a diagram showing a relationship between a propagation delay time and a maximum cell buffer queue length.

FIG. 15 is a diagram showing the concept of an ATM communication network based on the ABR method.

FIG. 16 is a diagram showing a processing flow of a conventional example.

FIG. 17 is a diagram showing a processing flow of a conventional example.

[Explanation of symbols]

1, 3, 4, 1-1, 1-2, 2-1 and 2-2 Switch 3-0, 3-1, 3-2, 3-3, 3-4 VP link 4-1 and 4-2 VC route 5-1 5-2 Sending end 6-1 Receiving end 11, 12, 21, 22, 41, 42, 51, 52 Route 31 Arrival cell counter 32 Arrival rate calculator 33 Input load factor calculator 34 Comparison Operation unit 35 RM cell transmission unit 36 RM cell extraction unit 37 Output rate control unit 38 Output rate calculation unit

Claims (4)

(57) [Claims] 1. An exchange accommodating a transmitting end and a receiving end.
And a physical transmission line connecting the exchanges to each other.
A relay exchange interposed in a physical transmission line.
VC (virtual channel) routes between exchanges
(Virtual Path) link that can be set
In the ATM communication network in which each is set, the exchange has its own input load factor (input rate / VP rate).
Means to calculate and record the input load factor.
As input load factor information for the exchange housed at the sending end
Transmission means and input load factor information coming from other exchanges
When there is information, the input load rate recorded by
Means for comparing the input load factor included in the force load factor information with
Wherein the sending means outputs the comparison result of the comparing means.
Therefore, the larger of the input load factors is
Including means for sendingSee The exchange accommodating the transmitting end may have a ratio of the comparing means.
According to the comparison result, Variable setting of VC route bandwidth set by own switch
With means to The means for variably setting the input load factor when it arrived last time
Includes means for estimating the current input load factor contained in the information
 An ATM communication network, characterized in that: (2)The means for estimating includes increasing and decreasing the output rate.
The ratio is γ, and the characteristics of this output rate increase / decrease ratio γ are determined.
Parameters are P1 and P2, and the output rate is R.
When the input load factor estimated value 率 pred is γ = P1 * Ψpred + P2  R = R * γ  Ψpred = Ψpred * γ Includes means to calculate as Claim1 noteATM communication
network. 3. The method of claim 1, wherein said estimated value and a newly arrived input negative
This estimate is based on the input load factor contained in the load factor information.
The AT according to claim 1 or 2, further comprising means for updating the AT.
M communication network. 4. The means for performing the update is newly arrived.
The input load factor that is included in the input load factor information Ψnotified, strange
The initial value of the number π is “1.0” and the value of this variable π is
The value of the variable π is sequentially obtained by multiplying by γ, and the updated
4. The ATM communication network according to claim 3 , further comprising means for calculating the estimated value Update @ prep as Update @ prep = min (@pred, $ notified * π) .