JP3197254B2 - ATM virtual path capacity setting method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an asynchronous transfer mode (ATM) network.
The present invention relates to an ATM virtual path capacity setting method for setting the capacity of a TM virtual path so as to satisfy a predetermined cell loss rate target value for the ATM virtual path.・ Related to technology for setting capacity in accordance with high-precision traffic demand.

[0002]

2. Description of the Related Art As is well known, a circuit-switched communication network such as a telephone service uses a synchronous transfer mode (STM;
Synchronous Transfer Mode). ST
In the M network, a line having a fixed band is set in a path set between two exchanges.
Only line setting with a minimum bandwidth of 64 kbps is possible,
Dynamic bandwidth change after line setting is impossible except for some terminals, and the minimum unit is 64 even for terminals capable of dynamic bandwidth change.
kbps. Even if a communication network that handles only fixed bandwidth lines is individually operated in this way, it is not possible to flexibly support various communication services and new communication services, and there is a limit to efficient operation of network resources. there were.

A data communication network such as a packet communication service is provided by packet communication which is a known technique. In a packet communication network, data from an unspecified number of information sources is converted into packets of various lengths and transferred, but the data from the specified information source affects all other data flowing through the same transmission medium. As a result, the utilization efficiency of the transmission medium cannot be sufficiently increased while keeping the data loss rate low. In this way, even when operating a communication network in which the bandwidth of the transmission medium is shared by unspecified large numbers without restriction, it is not possible to flexibly respond to various data characteristics and fluctuations in demand, and there is a limit to the efficient operation of network resources. was there.

On the other hand, in a recent wideband ISDN that handles a plurality of communication services such as telephone, data communication, and image communication in an integrated manner, an asynchronous transfer mode (ATM) is used.
With the advent of, by converting all communication information into fixed-length cells and transferring them, it becomes possible to realize a centralized exchange processing independent of the type of communication service, and instead of a path, two exchanges can be implemented. The concept of a virtual path, which is a logical path that specifies a usable band in the above, has been proposed, and it has become possible to logically set virtual circuits having various bands in the virtual path. Also, under the limitation of the capacity of the virtual path, it is possible for an unspecified number of information sources to logically set a virtual line sharing the virtual path.

[0005] Information transfer by a cell is a technique close to time-division multiplexing using a time slot, which is a known technique.
Time slots for virtual circuits are not allocated periodically, and time slots can be dynamically allocated in response to a time-varying information transmission request. Therefore, various bands can be variably set by changing the number of transmission cells per unit time. That is, in the ATM, all communication services from voice and data to images can be transmitted in a centralized manner.
A communication network that performs communication based on this ATM is called an ATM network.

[0006] Regarding the configuration of a transmission path network using virtual paths, for example, Sato, Kaneda and Tokizawa, "Configuration of High-Speed Burst Multiplexing Transmission System" (Institute of Electronics, Information and Communication Engineers Information and Communication Network Research Group, IN 87-) 84, 1987).

FIG. 1 illustrates a virtual path terminating at an ATM exchange and its cell transmission waiting buffer. First, the physical connection configuration will be described. The cell transmission device 100 of the ATM exchange has a cell transmission interface 101 for each destination physical line, and suppresses cell transmission exceeding the capacity of the physical line. Here, it is limited by the capacity of the physical line 200.

Next, a logical connection configuration will be described. The cell transmission interface has a shaper (sh
aper) 111 and 112 are connected, and the shaper suppresses cell transmission exceeding the capacity logically determined for the virtual path. Cells are transmitted between the cell transmission interface and the shaper at a timing according to a certain scheduling rule. Shapers and virtual paths correspond one-to-one, and shapers 111 and 112 correspond to virtual paths 201 and 202, respectively. For simplicity, the capacity of the virtual path is equated to the shaper's cell transmission speed. A cell transmission waiting buffer 121 or 122 is connected to the shaper. When a cell exceeding the virtual path capacity flows into the buffer, the buffer holds the cell and waits for the next cell transmission timing.

Next, a cell transmission waiting buffer, a Q length, a shaper, and a cell transmission interface will be described along the flow of cells. The cells switched according to the destination physical line in the ATM exchange are routed to the cell transmission device 100. Logically, the virtual path 20
Cells are divided for each of destinations 1 and 202, and incoming cell flows 301 and 302 are added to the cell transmission device 100, respectively. The buffer 121 to which the cell stream 301 is added counts the number of cells arriving during a predetermined period T.

Cells arriving at the cell transmission waiting buffer 121 are processed in the order of arrival. If the shaper 111 is in operation, it is accumulated in the buffer and waits for the turn of processing. The number of cells waiting here is the Q length. The order of cell processing in the buffer is based on a scheduling rule defined by the shaper. When the order of the processing comes to the shaper 111, the cells stored in the buffer 121 pass through the shaper 111, advance to the cell transmission interface 101, and are transmitted to the physical line 200.

Next, some conventional ATM virtual path capacity setting methods in the above-described ATM exchange will be described.

[Conventional method 1] First, as conventional method 1,
There is a method of modeling a cell arrival process by a Poisson process. This method has a feature that the virtual path capacity can be easily determined by only one traffic measurement item called the virtual path usage rate. That is, the buffer 12
1 is K, the cell loss rate design value is CLR, and the cell flow 301 with the arrival rate λ to the virtual path 201 is C
The virtual path capacity Cd that satisfies LR is calculated by the following equation (1). Cd = λM / ρ (1)

However, λ will be described in the following equation (7). M is a constant for converting the number of cells per unit time into capacity, and the usage rate ρ is given by the following equation (2). ρ = CLR ^{1 / (K + 1)} (2)

The above equation (1) is derived on the basis of the following. That is, if the Q length distribution is S and the probability that the Q length is longer than k is P [S> k], it is known that the following equation (3) is established by the large deviation theory. P [S> k] ≒ exp (−ρk) (3)

In particular, since it is assumed that the cell arrives according to the Poisson process, the following equation (4) holds.

(Equation 3)

Here, when Molina's formula is applied, the above equation (4) is expressed as an approximate equation (5). P [S> k] ≒ ρ ^{K + 1} (5)

This discussion relates to the M / D / s queuing model, and is described in detail in pages 351 to 362 of Communication Traffic Theory (Fujiki, Ganbe). In the above equation (5), when the Q length k on the left side matches the buffer size K, the cell discard rate design value CLR
, And the usage rate ρ at that time is obtained,
Equation (6) is obtained. CLR = ρ ^{K + 1} (6)

Next, FIG. 2 is a block diagram showing a functional configuration of an ATM exchange for realizing such a conventional method 1. As shown in FIG. In the following description, the same reference numerals are given to the components already described with reference to FIG. 1, and the description thereof will be omitted. In this ATM switch, known soft counters 131 and 132 are attached to buffers 121 and 122,
When cells are logically allocated to the virtual paths 201 and 202, the number of arrival cells for each virtual path is counted.

The arrival rate λ is determined by the soft counters 131 and 13.
In the setting device 400 that summarizes the counting results of 2, the arrival rate λ, which is the number of cells arriving per unit time, is calculated by the following equation (7) for each predetermined period T. λ = N / T (7)

However, as a general property of the ATM cell flow, there is no guarantee that modeling of the cell flow by the Poisson process is sufficiently valid. That is, in the ATM network, the burstiness of traffic cannot be said to be small enough to be modeled in the Poisson process. Reports such as large burstiness and long-term correlation with respect to ATM traffic are described in, for example, W.
`` On the self-similer nature of Ethernet traff '' by E. Leland, MS Taqqu, W. Willinger, and DV Wilson
ic (extended version) ”IEEE / ACM Trans.Networking,
vo1.2, no.1, pp1-15,1994. Therefore, designing the virtual path capacity of the ATM network by the conventional method 1 has a problem that traffic overflow frequently occurs and information is easily lost.

[Conventional method 2] Next, as conventional method 2,
A method of calculating the virtual path capacity with no excess and shortage by collecting all the arrival times of the cells and restoring the arrival process will be described. FIG. 3 shows A for realizing the conventional method 2.
It is an example of a functional configuration of a TM exchange. In this ATM exchange, an external device 500 is inserted between the cell sending interface 101 and the physical interface 200 to copy (capture) a flowing cell, or copy a header part which is a part of the cell, and to copy the header through a communication line 501. Storage device 5
10, the copied data is stored.

For example, when a cell arrives, the storage device 510 stores the time stamp of the arrival time and the header of the cell. The header of this cell contains an identifier for specifying the virtual path. The storage device 510 classifies these timestamps for each virtual path and analyzes the cell arrival process for each virtual path. As the analysis method, there is a method of creating a probability distribution of an arrival process, a method of calculating an average or variance of an arrival interval, a third moment, or the like.

However, the storage device 510 capable of giving such a time stamp is expensive. The main cause is
This is due to the technical difficulty of time-stamping in sub-microsecond units required to analyze incoming cells. In addition, continuous capture can be performed only while the amount of memory mounted in the storage device permits (capture for several seconds is performed every few minutes), so that design accuracy at intervals other than the capture interval cannot be guaranteed.

Further, depending on the specifications of the external device, some devices cannot capture only the header information added by the communication provider, and there is a possibility that the payload carrying the customer's communication information may be sent to the outside. You have to be careful. Further, communication disconnection when connecting the external device 500 between the cell transmission interface 101 and the physical line 200 is inevitable. In other words, in this conventional technique, in order to design the virtual path capacity,
This will hinder customer communication.

Further, since the cell flow that can be captured is a cell flow prepared by the shaper, the burst property will be small, and a cell lost in the buffer will not be captured. This has a small effect on low-load operation, but has a large effect on the accuracy of virtual path capacity design during high-load operation. Therefore, in designing the virtual path capacity of the ATM network based on the conventional method 2, it is necessary to solve the above-mentioned problem.

[Conventional Method 3] Next, prior art 3 will be described, but before that, the Q length distribution shown in FIG. 4 will be described. This Q length distribution is created from the traffic data actually measured by the same method as the conventional method 2 described above.
The Q length distribution refers to the number of cells (Q
Long) probability distribution. In the graph shown in FIG. 4, the horizontal axis represents the number k of cells in the buffer, and the vertical axis represents the probability P [S> k] that cells equal to or more than the number of cells are stored in the buffer in logarithm. Each dashed line in the graph represents a Q length distribution based on arrival time data of about 130,000 cell traffics. Further, FIG. 4 shows a Q length distribution around a straight line approximating the Q length distribution (a portion surrounded by a dotted line in FIG. 4).

As shown in FIG. 4, the Q length distribution becomes lower rightward from the origin at the upper left of the graph as the Q length k increases, and the absolute value of the slope of this graph is the attenuation rate.
As a general feature of the Q length distribution, the attenuation rate gradually decreases as the distance from the origin increases (the Q length k increases), but there is a portion where the decrease becomes slower when a certain Q length value is reached. The portion where the decrease becomes slow is called the tail of the Q length distribution.

Normally, in a range smaller than the reciprocal of the number of digits of the measured number of cells, the Q length distribution has no meaning. For example, as shown in FIG. 4, traffic data obtained by capturing 130,000 cells is used. Then, a probability of less than 10 ^-4 is not sufficient.

In general, it is known that, in a stochastic process that satisfies stationarity, rarity, and independence, the probability of occurrence of an event follows an exponential distribution (the Poisson's law of a small number). this is,
In the case of the system configuration of FIG. 2, it means that the tail of the Q length distribution is approximated by a straight line. Here, the Q length threshold value and the number of times the Q length threshold value is exceeded will be described. As described in the conventional method 2, it is technically difficult to grasp the distribution of the Q length.

In order to partially understand the characteristics of the Q length, a mark (Q length threshold) is provided in the middle of the buffers arranged in a line.
Each time a cell arrives, it is determined whether or not the Q length has exceeded the Q length threshold, and the result is counted, and the count value (the number of times the Q length threshold has been exceeded) is determined at predetermined intervals. There is a measurement technique that outputs Unlike the method of capturing the cell arrival process described in the conventional method 2, this is easy to implement because it can be attached to the existing process of waiting for cell transmission, and the number of data is periodically output. Thus, the data can be compressed for the measurement period, and continuous measurement can be performed by data compression. In addition, the connection can be counted regardless of the customer's payload, and the measurement position can be measured before the shaper and before the rectification by the shaper. Note that a value obtained by normalizing the number of times the Q length threshold is exceeded by the number of arrival cells is the Q length threshold excess frequency.

Conventional method 3 is characterized in that two points of Q length are used by using the above-mentioned technique, and the Q length distribution is obtained by a straight line determined from two Q length threshold excess frequencies obtained therefrom. This is a method to approximate the tail. The solid line shown in FIG. 4 indicates that the Q-length distribution is approximated by a straight line passing through the Q-length threshold at a certain two points and the corresponding Q-length threshold excess frequency.

Hereinafter, referring to the functional configuration diagram of FIG. 5, the case where two Q length threshold values are set in each cell transmission waiting buffer 121 or 122 of the cell transmission device 100 (in FIG. 5,
(Shown by “▲” and “▼”) will be described. The first Q length threshold 151 (1) and the second Q length threshold 151 are stored in the cell transmission waiting buffers 121 and 122, respectively.
(2) The first Q length threshold 152 (1) and the second Q length threshold 152 (2) are set, and when the cell arrives, the Q length exceeds the Q length threshold. Is determined. Of the determination results, the Q length is the first and second Q
The result exceeding the long threshold value is counted at the same cycle T as the counting of the number of arriving cells. The counting result is the number of times the Q length threshold has been exceeded.

FIG. 6 shows a table in which data of a part relating to the virtual path 201 in the counting result of the cell transmission device 100 is summarized. The data counted by the cell sending device 100 is sent to the setting device 400 (see FIG. 5), and is stored in the setting device 400 as a table shown in FIG. This table records the number of arrival cells, the number of times the first Q-length threshold is exceeded, and the number of times the second Q-length threshold is exceeded, for each predetermined period T. That is, in the table of FIG. 6, the number of the counting cycle is recorded in the column of “counting cycle T”, and the number N of arrival cells corresponding to the number of the counting cycle, the number of times the first Q length threshold is exceeded, Q ₁ and the second Q-length threshold excess count Q ₂ are the “arrived cell count”, “first Q-length threshold exceed count”, and “second Q It is recorded in each column of "number of times of long threshold exceeded".

The setting device 400 processes the data in the table shown in FIG. 6 to calculate the design capacity for the virtual path 201. Note that this calculation formula is based on Shioda, Toy
oizumi, Yokoi, Tsuchiya, Saito, “Self-sizing network:
a new network concept based on autonomous VP bandw
idth adjustment, ”Proc.of ITC 15, pp.997-1006,1997
It is described in.

First, using the record N of the number of arrival cells in the table, the virtual path usage rate ρ is calculated by the following equation (8). ρ = MN / C (8) where M is a constant for converting the number of cells per unit time into capacity, and C is the capacity set for the virtual path 201.

Further, a record Q _{1 of} the first Q length threshold exceeded count and a record Q ₁ of the second Q length threshold exceeded count
₂ , the first threshold value exceeding frequency P _{1 is obtained} by the equation (9).
After calculating the second and threshold crossing frequency P _2. P ₁ = Q ₁ / N, P ₂ = Q ₂ / N (9)

The meaning of "frequency" herein referred to _one or Q ₂ or Q of the N cells arriving during the measurement period T is caused a phenomenon that exceeds a threshold value corresponding to each Probability. That is, in a finite trial experiment such as rolling a dice, this has the same meaning as the probability Q / N obtained when a specific event occurs Q times after performing N times during the T time.

Further, the Q length distribution attenuation rate δ is calculated by the following equation (10).
It is calculated by: Here, k ₁ is the first Q length threshold 1
51 (1), and, Q ₂ is the second Q length threshold 151 (2) (see FIG. 5).

(Equation 4)

The attenuation factor β is calculated by the following equation (11).

(Equation 5)

Next, the required band Cd is calculated by equation (12).

(Equation 6) Here, m ⁽¹⁾ is as shown in the following equation (13), and m ⁽²⁾
Is assumed to be as shown in (14). Also, T
Is the length of the counting cycle, K is the buffer size, and CLR is the cell loss rate design value.

M ⁽¹⁾ = δlogβ (13) m ⁽²⁾ = (ρ / (1−ρ)) (1 / δ) (14)

In equation (12), the required bandwidth is calculated using the data of the number N of arriving cells at a specific point in time. For example, by using the following equations (15) and (16), The conventional method 3 can also be applied when N data is obtained in a time series. Cd = Max {Ci | i = 1,..., N} (15) Cd = Cd (n), Cd (i + 1) = (1−α) Cd (i) + αCi, i = 1,. 1 (16)

Here, α is a constant satisfying 0 ≦ α ≦ 1. C
i means the required band obtained from only the data of the i-th measurement cycle, and Cd (i) means the required band obtained from the data of the first to i-th measurement cycles.

The hypothesis that equations (13) and (14) are connected by an equal sign, respectively, is within a range where the calculation does not have an effect (eg, 1 Mbps or less) that has no practical effect on calculating the required bandwidth. However, depending on the traffic characteristics, the range in which the equal sign is established is narrowed, and the accuracy of the calculated required band is reduced.

[Conventional Method 4] Next, a conventional method 4 will be described with reference to FIGS. 7 and 8. This is Japanese Patent Application No. 09-2
In this method, one Q length threshold value is set by the method described in detail in the embodiment of No. 85845, "ATM switch and ATM virtual path capacity setting method".

Hereinafter, referring to the functional configuration diagram of FIG. 7, each cell transmission waiting buffer 121 or 12 of the cell transmission device 100 will be described.
In the case where one Q-length threshold value is set in FIG.
(Indicated by a “▼” mark). Each of the cell transmission waiting buffers 121 and 122 has a Q length threshold value of 1
41 and 142 are set respectively, and it is determined whether or not the Q length exceeds the Q length threshold when a cell arrives. Among the determination results, the results in which the Q length exceeds the Q length threshold value are counted at the same cycle T as the counting of the number of arriving cells. The counting result is the number of times the Q length threshold has been exceeded.

FIG. 8 shows a table in which data of a part relating to the virtual path 201 in the counting result of the cell transmission device 100 is summarized. The data counted by the cell transmitting device 100 is transmitted to the setting device 400 (see FIG. 7), and is stored in the setting device 400 as a table shown in FIG. This table records the number of arriving cells and the number of times the Q length threshold has been exceeded for each predetermined cycle T. That is, in the table of FIG. 8, the number of the counting cycle is recorded in the column of “counting cycle T”, and the number N of arriving cells and the number Q of exceeding the Q length threshold value corresponding to the number of the counting cycle are: They are respectively recorded in the columns of “number of cells arriving” and “number of times the Q length threshold has been exceeded” in the table. Then, the setting device 400 processes the data in the table shown in FIG. 8 and calculates the design capacity for the virtual path 201.

First, using the record N of the number of arrival cells in the table, the usage rate ρ of the virtual path is calculated by the following equation (17). ρ = MN / C (17)

Here, M is a constant for converting the number of cells per unit time into capacity, and C is the capacity set for the virtual path 201. Based on equation (18), the Q length distribution attenuation rate δ ₀ when the cell arrival follows the Poisson process is calculated. δ ₀ = −logρ (18)

The record Q indicating the number of times the Q length threshold value has been exceeded
Is used, the Q length threshold value excess frequency P is calculated based on the equation (19). P = Q / N (19)

Further, the calculated Q length threshold value excess frequency P
Estimated value δ of Q length distribution decay rate based on ₁ Becomes the following equation (20)
Therefore, it is calculated. δ₁=-(1 / k) log P (20)

In this equation (20), k is a Q length threshold value set in the transmission waiting buffer 121.
Then, the ratio w of the Q length distribution attenuation rate calculated based on the above equations (18) and (20) is calculated by the following equation (21). w = δ ₀ / δ ₁ (21)

Using the above calculation results, the required bandwidth Cd of the virtual path capacity to be obtained is calculated by equation (22).

(Equation 7) Here, CLR is a design value of the cell loss rate, and K is a buffer size.

In the equation (22), the required band is calculated using the data of the number N of arriving cells at a specific point in time. For example, w and d are calculated using the following equations (23) and (24).
, The conventional method 4 can be applied even when data of the number N of arriving cells is obtained in time series. w = Max {wi | i = 1,2, ..., n} (23) N = Max {Ni | i = 1,2, ..., n} (24) where the subscript i is the i-th measurement cycle Means the data.

[0055]

The ATM described above
The related art relating to the setting of the virtual path capacity has the following problems.

(1) In the conventional method 1, the ATM exchange can determine the parameters for calculating the virtual path capacity with simple traffic measurement items. However, the ATM cell traffic characteristics are only expected to fluctuate in the Poisson process, and There is a problem that it is not possible to take into account the nature and long-term correlation.

(2) In the conventional method 2, it is possible to accurately calculate the virtual path capacity by capturing the cell traffic data by the external device. However, in general, the external device is expensive, and the external device is expensive. The operation is also complicated. In addition, while connecting the external device, the communication of the customer must be stopped, the captured data needs to be separately analyzed, and the obtained analysis result can be applied only in a short time during which the capture can be performed. There is a problem that there is.

(3) In the conventional method 3, it is possible to obtain a straight line that approximates the attenuation rate of the tail of the Q-length distribution relatively accurately from the two Q-length threshold excess frequencies, and thereby the virtual path capacity. Can be calculated. However, there is a problem that the application range of this method is limited and cannot be widely applied.

(4) In the conventional method 4, the approximate straight line can be easily obtained by estimating the decay rate of the tail of the Q length distribution from one Q length threshold value exceeding frequency. However, there is a problem that the calculation accuracy fluctuates according to the burstiness of traffic.

The present invention has been made in view of the above circumstances, and requires a special measurement form such as the above-described external device by using data obtained by simple traffic measurement in an ATM exchange. In addition, an object of the present invention is to provide an ATM virtual path capacity setting method which can set a virtual path capacity of an ATM network and can obtain higher calculation accuracy in a wide range.

[0061]

In order to achieve the above object, according to the present invention, there is provided an ATM virtual path capacity setting method for setting a virtual path capacity satisfying a cell discard rate design value for an ATM virtual path terminating at an ATM exchange. At the ATM
Two or more Q-length thresholds are set in the cell transmission waiting buffer associated with the virtual path, and the number of cells arriving at the cell transmission waiting buffer and waiting in the cell transmission waiting buffer at predetermined intervals. , The number of times the number of cells exceeds the set Q length threshold, and the tail of the Q length distribution is approximately obtained by a function showing a straight line, and the constant with respect to the band set in the ATM virtual path is determined. Calculate the bandwidth usage rate which is the ratio of the number of cells arriving per cycle, and calculate the bandwidth usage rate, the cell discard rate design value, the tail of the Q length distribution obtained approximately, and the cell transmission waiting buffer. Calculating a target usage rate based on the buffer size;
It is characterized in that a minimum required bandwidth of the virtual path is calculated based on the fixed period and the number of arriving cells to satisfy the cell loss rate design value.

Here, as described above, the tail of the Q length distribution refers to, for example, a portion where the decrease of the attenuation rate is slowed down in the graph of the Q length distribution as shown in FIG.

In the present invention, the cell loss rate design value is
The specific calculation of the required bandwidth, which is the minimum required bandwidth of the virtual path to satisfy the condition, is as follows. First, assuming that the calculated bandwidth usage rate is U _R , a predetermined constant cycle is T, Assuming that the number of cells (the number of arriving cells) arriving at the cell transmission waiting buffer during this fixed period is N and the bandwidth set for the ATM virtual path is C, the bandwidth usage rate is expressed by the following equation: U _R = N / (TC) Calculate U _R.

When the target utilization rate to be calculated is R, the tail of the Q length distribution is represented by a straight line represented by a function F (k) = α exp (−β k) using the Q length threshold k as a variable. Using the intercept constant α and the slope (ie, attenuation rate) β of the function F (k) and the calculated bandwidth utilization rate U _R , further, the cell loss rate design value is CLR, Assuming that the buffer size of the transmission waiting buffer is K, the target usage rate R is calculated from the following equation.

[0065]

(Equation 8)

Then, from the calculated target utilization rate R, the fixed period T, and the number N of arriving cells , Cd, which is the required band finally obtained , is calculated using the following equation. Cd = N / (TR)

Here, the above-mentioned intercept constant α and inclination β
(Hereinafter referred to as attenuation rate β) is calculated as follows.
First, two Q length thresholds k ₁ and k ₂ (k ₁
Select <k ₂ ). The number of times that the number of cells waiting in the cell transmission waiting buffer exceeds the selected Q length thresholds k ₁ and k ₂ during the predetermined period T (the number of times the Q length threshold is exceeded). each and Q _1, Q ₂ seeking. Then, the constants α and β are calculated from the following equations using the Q length thresholds k ₁ and k ₂ , the Q length threshold value excess times Q ₁ and Q ₂ , and the number N of arrival cells. β = −1 / (k ₂ −k ₁ ) · log (Q ₂ / Q ₁ ) α = (Q ₁ / N) exp (βk ₁ )

In the above-mentioned calculation of the required band,
When cell discarding occurs in the cell transmission waiting buffer, the number of discarded cells may be set to L, and the above-mentioned attenuation rate β may be calculated using the following equation instead of the above-mentioned equation.

(Equation 9)

Further, the selected two Q length threshold values are
That is, the larger Q length threshold value k _Two Q length for
Value overrun Q_Two Is smaller than the predetermined lower limit.
The intercept constant α and the attenuation rate β, respectively.
Β = −1 / k₁・ Log (Q₁/ N) α = 1.

[0070] Between the next larger Q length threshold k ₁ and Q length threshold k ₂ above, so as to have a difference of more than 5 cells, determine the Q length threshold k ₁ You may do so.

According to the present invention, it is possible to set the capacity of the ATM virtual path in the ATM network at high speed and with high accuracy in accordance with the actual traffic demand based on the light traffic measurement. The first feature is that in the traffic measurement at the cell level, the virtual cell discard rate at the Q length threshold is grasped in spite of the simple measurement of comparing the Q length threshold and the Q length. Is a point. As a result, the behavior of the cells in the buffer waiting to be transmitted can be grasped relatively easily and accurately, and the AT corresponding to the traffic demand can be determined.
The capacity of the M virtual paths can be set with high speed and high accuracy.

The second feature is that the tail of the Q length distribution is estimated by measuring the number of times the Q length threshold is exceeded at two points in the behavior of the Q length in the cell level traffic measurement. Thereby, the Q length distribution in the transmission waiting buffer can be grasped relatively easily, and the capacity of the ATM virtual path according to the traffic demand can be set with high speed and high accuracy.

A third feature is that a required band can be calculated over a wide range with a relatively high degree of accuracy.
Thus, the difference between the traffic demand and the calculated required bandwidth can be reduced.

With these characteristics, in particular, UBR (Unspec
Unified Bit Rate) Set the capacity of the ATM virtual path that accommodates the best-effort virtual circuit including the bearer class so as to satisfy a certain cell loss rate design value (not a specified value for service provision to customers). It becomes possible to do.

Furthermore, it is possible to check and record daily that the resources of the ATM network are appropriately arranged in the ATM virtual path and that they are being used effectively, and that the traffic flows evenly throughout the network. Confirmation becomes possible. In addition, it is possible to provide information for performing appropriate resetting in a reasonable operation and time with respect to a network configuration change such as a new addition or relocation of a virtual circuit.

[0076]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of an ATM virtual path capacity setting method according to the present invention will be described below with reference to the drawings and equations. In the following description, the same reference numerals are given to the components already described, and the description thereof will be omitted. In particular, the functional configuration in the following embodiment is the same as the functional configuration diagram of FIG.

[First Embodiment] A first embodiment of the present invention will be described with reference to FIGS. First,
The function configuration and the record of the data stored in the setting device 400 are the same as in the conventional method 3. Next, a calculation method will be described. First, virtual paths 201, 202,.
The ratio (band use rate) U _R of the number N of cells flowing per measurement period T to the band C set in each of the... Is calculated by the following equation (25). U _R = N / (TC) (25)

Next, estimation of the Q length distribution will be described. As described in the conventional method 3, by using two Q length thresholds, the tail of the Q length distribution is represented by a straight line F (k) = α exp (−β
k) can be approximated. Since two Q length threshold value excess counts have been acquired, Q ₁ / N at k = k ₁
And the value of Q ₂ / N at k = k ₂ are represented by F
(k) = α exp (−βk). As a result, a two-dimensional simultaneous linear equation of α and β is obtained. Solving this,
Equations (26) and (27) are obtained. β = −1 / (k ₂ −k ₁ ) · log (Q ₂ / Q ₁ ) (26) α = (Q ₁ / N) exp (βk ₁ ) (27)

Up to the calculation of the approximate expression, the conventional method 3 and the first embodiment are the same, but the calculation method of the required band is different. In order to explain the method of calculating the required bandwidth according to the first embodiment, the maximum usage rate (target usage rate) R that satisfies the cell loss rate design value will be described. If the cell is to arrive according to a Poisson process, the attenuation rate of the hem of Q length distribution β _{Poisso n}
Is known to follow the following equation (28). β _Poisson = -logρ (28)

Where ρ is the cell usage rate for the band of the virtual path. Numerically, U _R = ρ, but since the characteristics of the arrival process are not always the same, the description will be made with different symbols. The target usage rate R _{Poisson at} which the cell discard rate satisfies the cell discard rate design value CLR is expressed by equation (29) by the slope of a straight line passing through the origin shown in FIG. −log (CLR) / K = −log (R _Poisson ) (29)

When this equation is solved for R _Poisson , equation (30) is obtained.

(Equation 10)

As a result, the required band Cd _Poisson is _calculated by the equation (3)
1). Cd _Poisson = N / (TR _Poisson ) (31) where T is the measurement time and the number of cells arriving during that time is N
It is.

Even if the cell arrival does not follow the Poisson process, if there is a relational expression between the current usage rate U _R and the decay rate β at the tail of the Q length distribution as shown in equation (28), the Poisson process Calculate the target usage rate R when not complying with
d can be calculated. Therefore, a hypothesis that the following equation (32) holds using the parameter γ was adopted. β = −γlogρ (32)

With respect to this hypothesis, it was verified that a certain degree of accuracy was established for a wide range of usage rates by measuring the detailed traffic data described in the conventional method 2. For example, the attenuation rate β of the Q length distribution shown in FIG. 4 is slower than the attenuation rate β _{Poisson of the} Poisson process obtained from the usage rate ρ. To give specific γ, equation (32) may be solved for γ. Although the value of γ varies depending on the traffic characteristics, it has been confirmed that the form of Expression (32) can be established over a wide range with a certain degree of accuracy using various traffic data.

From U _{R in} equation (25) and β in equation (26),
The following equation (33) holds. β = −γ log (U _R ) (33)

The following equation (34) holds from the attenuation factor β _Object of the Q-length distribution when cells flow at the target usage rate R. β _Object = -γlogR (34)

Eliminating γ from these two equations gives log R
Solving for gives equation (35). _{logR = (β Object / β)} · log (U R) (35)

The value of β _Object can be obtained from the graph of the approximate straight line at the tail of the Q length distribution shown in FIG.

[Equation 11]

By substituting this into Expression (35) and solving for the target usage rate R, Expression (37) is obtained.

(Equation 12)

Based on the value obtained by equation (37), the required band C
d is obtained by equation (38). Cd = N / (TR) (38)

[Second Embodiment] Next, a second embodiment of the present invention will be described. In the first embodiment,
Using Q length threshold Q _1, Q length threshold number of times of excess is virtually number dropped cells with Q ₂ Q _1, Q _2, Q
While the method for obtaining the attenuation factor β at the tail of the length distribution has been described, the second embodiment describes a method that is not virtual but is applied when a cell is actually discarded.

In this embodiment, the Q length threshold k
₁ , k ₂ and the buffer size K, the number Q ₁ , Q ₂ exceeding the Q-length threshold, which is the number of cells virtually discarded,
A method of obtaining the attenuation rate β at the tail of the Q length distribution using the number of cells actually discarded, that is, the number L of discarded cells will be described.

[0093] Since the virtual cell discard rate and the actual cell discards are obtained three points, firstly, Q ₁ in the k = k ₁
/ N and the value of Q ₂ / N at k = k ₂ are respectively substituted for F (k) = α exp (−β k), and further, k = k
The value of Q ₂ / N at _2, the value of L / N at k = K, respectively substituted into F (k) = α exp ( -βk). As a result, two sets of binary first-order simultaneous equations relating to α and β are obtained. Β is obtained for each of the two sets in the same manner as in the first embodiment. In order to derive a safe design value, a solution with a smaller value among the two solutions of β is used in subsequent calculations. The solution with the small value is obtained by equation (39).

(Equation 13)

Hereinafter, using the value of β obtained by equation (39), equations (37) and (38) are used in the same manner as in the first embodiment.
To obtain the required band Cd.

[Third Embodiment] Next, a third embodiment of the present invention will be described. In the first embodiment, Q
Using the long thresholds k ₁ , k ₂ and the Q-length threshold excess times Q ₁ , Q ₂ for each, the decay rate β of the tail of the Q-length distribution was determined. In the embodiment,
The case can not use the Q-length threshold crossing times Q ₂ to Q-length threshold k ₂ larger, describes a method for determining the β attenuation rate of the hem of the judgment and Q length distribution.

Q length threshold for large Q length threshold
The number of excess times tends to be smaller. For example,
Q length is k_TwoIf not reached, Q_TwoIs zero. Q long threshold
If the number of excess values is too small,
Accuracy also decreases. Therefore, regarding the number of times the Q length threshold has been exceeded,
Threshold is determined in advance, and if it is smaller than that, k _Two 
And Q_TwoA calculation formula that does not use / N is used. The calculation formula is
(40) and equation (41). β = -1 / k₁・ Log (Q₁/ N) (40) α = 1 (41)

As the predetermined threshold value, the latter half of a 2-digit natural number (any one of 51 to 99; for example, 6
4), the accuracy of the calculation result is also about two digits, which is sufficient accuracy as a design value of the virtual path capacity.

In each of the embodiments described above, even when three or more Q length thresholds are set, the application is easy by adopting the second to the largest Q length thresholds. It is possible. When applying each of the above-described embodiments to time-series data obtained periodically,
For example, by calculating a statistic such as Expression (15) or Expression (16), it is possible to easily reduce the number of times of setting the capacity and reduce the number of operations.

[0099]

As described in detail above, according to the present invention,
In an ATM virtual path capacity setting method for setting a virtual path capacity that satisfies a cell discard rate design value for an ATM virtual path terminating in an ATM exchange, two Q length thresholds are set in a cell transmission waiting buffer associated with the ATM virtual path. With the above settings, the number of cells arriving at the cell transmission waiting buffer and the number of times that the number of cells waiting in the cell transmission waiting buffer exceed the set Q length threshold value are determined at predetermined intervals. Counting, the tail of the Q length distribution is approximately obtained by a function indicating a straight line, and a band usage rate which is a ratio of the number of cells arriving per the fixed period to the band set in the ATM virtual path is calculated. The band usage rate, the cell loss rate design value, the tail of the Q length distribution obtained approximately, and
A target usage rate is calculated based on the buffer size of the cell transmission waiting buffer, and the minimum required to satisfy the cell discard rate design value based on the target usage rate, the fixed period, and the number of arriving cells. Since the bandwidth of the virtual path is calculated, it is possible to efficiently estimate the cell discard in the cell transmission waiting buffer associated with the ATM virtual path, and set the ATM virtual path capacity according to the traffic demand with high speed and high accuracy. Further, even when cell discarding actually occurs or when the value of the number of times the Q length threshold has been exceeded is small, a decrease in accuracy can be suppressed.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a partial functional configuration of a conventional ATM exchange.

FIG. 2 is a block diagram showing a functional configuration of a part of an ATM switch for realizing a conventional method 1 relating to an ATM virtual path capacity setting method.

FIG. 3 is a block diagram showing a functional configuration of a part of an ATM switch for realizing Conventional Method 2 relating to an ATM virtual path capacity setting method.

FIG. 4 is a diagram illustrating a graph for explaining a conventional method 3 relating to an ATM virtual path capacity setting method.

FIG. 5 is a block diagram showing a partial functional configuration of an ATM exchange for implementing Conventional Method 3 relating to an ATM virtual path capacity setting method.

FIG. 6 is a diagram showing a table of data stored in a setting device in an ATM exchange for implementing the conventional method 3 relating to the ATM virtual path capacity setting method.

FIG. 7 is a block diagram showing a functional configuration of a part of an ATM exchange for realizing a conventional method 4 relating to an ATM virtual path capacity setting method.

FIG. 8 is a diagram showing a table of data stored in a setting device in an ATM exchange for realizing a conventional method 4 relating to an ATM virtual path capacity setting method.

FIG. 9 is a graph showing a Q-length distribution and a straight line for estimating the attenuation rate of the tail thereof in the first embodiment of the present invention.

[Explanation of symbols]

 REFERENCE SIGNS LIST 100 Cell transmission device 101 Cell transmission interface 111, 112 Shaper 121, 122 Cell transmission wait buffer 200 Physical line 201, 202 Virtual path 301, 302 Cell flow 131, 132 Soft counter 400 Setting device 401, 501 Communication line 500 External device 510 Storage apparatus

────────────────────────────────────────────────── ─── Continuation of the front page (56) References JP-A-9-107368 (JP, A) JP-A-11-122267 (JP, A) JP-A-11-243398 (JP, A) IEICE Technology Research Report SS E99-16 (May 20, 1999) Hirofumi Yokoi et al., "Traffic Management of ATM Backbone Network", NTT R & D, Vol. 48, No. 9, pp. 683-687, September 10, 1999 Date (58) Field surveyed (Int.Cl. ⁷ , DB name) H04L 12/28 H04L 12/56

Claims (6)

(57) [Claims] 1. An ATM virtual path capacity setting method for setting a virtual path capacity that satisfies a cell discard rate design value for an ATM virtual path terminating at an ATM switch, wherein a Q length is added to a cell transmission waiting buffer associated with the ATM virtual path. By setting two or more thresholds, the number of cells arriving at the cell transmission waiting buffer and the number of cells waiting in the cell transmission waiting buffer are determined by the set Q length threshold value at predetermined intervals. The number of excess cells is counted, and the tail of the Q-length distribution is approximately determined by a function indicating a straight line. The band usage, which is the ratio of the number of cells arriving per fixed period to the band set in the ATM virtual path, is used. Calculating a rate, based on the band usage rate, the cell discard rate design value, the tail of the Q length distribution obtained approximately, and the buffer size of the cell transmission waiting buffer. Calculating a target usage rate, and calculating a minimum necessary bandwidth of the virtual path to satisfy the cell discard rate design value based on the target usage rate, the fixed period, and the number of arriving cells. A
TM virtual path capacity setting method. 2. Using the value of the number N of arriving cells, the fixed period T, and the value of the band C set in the ATM virtual path, the band use rate U _R is calculated as follows: U _R = N / (TC The intercept constant α and the slope β of the straight line approximating the tail of the Q length distribution
By using the intercept constant α and the slope β, the band use rate U _R , the cell discard rate design value CLR, and the value of the buffer size K, the target use rate R is calculated as follows: Using the calculated target utilization rate R, the constant period T, and the number of arriving cells N, the cell discard rate design value is calculated using the following formula:
2. A according to claim 1, wherein a required bandwidth Cd, which is a minimum required bandwidth of the virtual path to satisfy the condition, is calculated from an equation: Cd = N / (TR).
TM virtual path capacity setting method. 3. The set Q length thresholds k ₁ , k ₂ (k ₁
<K ₂ ), and the selected Q length thresholds k ₁ , k ₂ are set to the number of times that the number of cells waiting in the cell transmission waiting buffer during the predetermined period is exceeded. each Q _1, Q _2, and then, the number of arrival cells with N, said the intercept constants α and the slope of the straight line beta approximating the hem of Q length distribution, _{respectively, β = -1 / (k 2} -k _A ) according to claim 2, wherein the value is calculated from the following expressions: ₁ ) · log (Q ₂ / Q ₁ ) and α = (Q ₁ / N) exp (βk ₁ ).
TM virtual path capacity setting method. 4. When cell discarding occurs in a cell transmission waiting buffer, the number of discarded cells is L, and the slope β of a straight line approximating the tail of the Q length distribution is 4. A according to claim 3, wherein the value is obtained from the following equation:
TM virtual path capacity setting method. 5. A Q length threshold value exceeding number Q ₂ for a larger Q length threshold value k ₂ of the _{two selected} Q length threshold values is smaller than a predetermined lower limit value. In the case, the intercept constant α and the slope β of the straight line approximating the tail of the Q length distribution
The ATM virtual path capacity setting method according to claim 3 or 4, wherein β is calculated from the following expressions: β = −1 / k ₁ · log (Q ₁ / N) and α = 1. Between 6. next higher and the Q long threshold k ₂ Q length threshold k _1, so as to have a difference of more than 5 cells, to determine the Q length threshold k ₁ The ATM virtual path capacity setting method according to any one of claims 2 to 5, wherein: