JPH11355312A - Evaluation method for cell traffic characteristic via plural atm exchanges - Google Patents

Abstract

PROBLEM TO BE SOLVED: To evaluate an inter-end cell delay time distribution and the upper limit of cell loss in an ATM network where plural exchanges are connected in series by using a selected queue to decide a parameter for decision of the cell traffic characteristic and then deciding the inter-end queuing time distribution and the cell loss rate of the corresponding call. SOLUTION: When the control is started at a call reception control part 19, a route selecting function 16 decides the presence or absence of a connectable path. If a connectable path exists, a calculating parameter is decided based on the information received from a cell exchange 12 and notified to the reception control function of the part 19. The call reception control function uses the said parameter to decide an inter-end queuing time distribution and a cell loss rate. Furthermore, an additional process is carried out to decide whether the corresponding call satisfies the prescribed request quality. If the request quality is satisfied, the call is connected. If no connectable path exists, the connection of the corresponding call is rejected. Thus, a circuit is managed and a proper route is set.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a plurality of ATMs (As
The present invention relates to a method for evaluating cell traffic characteristics via an exchange, and particularly to a method for evaluating a delay time distribution and a cell loss rate.

[0002]

2. Description of the Related Art In a conventional study on the traffic characteristics via a plurality of ATM exchanges, background traffic affecting traffic of interest arrives at each exchange newly and independently and leaves the network after passing one link. [(Reference 1: Matteo D'Ambrosi
o and Riccardo Melen, "Evaluating the Limit Behavi
or of the ATM Traffic Within a Network ", IEEE ACM
Trans. Networking, vol. 3, No. 6, pp. 832-841, 199
5.); (Reference 2: Wassim Matragi, Khosrow Sohrab
y and Chatschik Bisdikian, "Jitter Calculus in ATM
Networks: MultipleNodes ", IEEE ACM Trans. Networki
ng, vol.5, No.1, pp.122-133, 1997.)].

[0003]

However, actual ATMs
In a network, it is often seen that the route of the traffic of interest and the route of the background traffic overlap over a plurality of links. In such a case, it is necessary to consider both the background traffic route and the load distribution. However, no evaluation method has been proposed for the traffic characteristics when the combination of the traffic route and the load distribution is variously changed.

The present invention shows a simulation result of traffic characteristics when a combination of a traffic route and a load distribution is changed, and considers a traffic behavior in each exchange, and a plurality of exchanges are connected in series. Multiple ATMs for evaluating end-to-end cell delay time distribution and upper limit of cell loss in ATM network
Provided is a method for evaluating cell traffic characteristics via an exchange.

[0005]

In order to solve this problem, a method for evaluating cell traffic characteristics via a plurality of ATM exchanges according to the present invention is provided for evaluating cell traffic characteristics via a plurality of ATM exchanges. A first step for determining a type of a cell arrival process of the cell traffic at a desired one of the plurality of ATM switches through which the call K _{i to} be evaluated passes; and A second step of selecting a queue suitable for obtaining according to the type of the cell arrival process defined in the above, and a third step of obtaining parameters necessary for obtaining the cell traffic characteristics using the selected queue. a step, waiting time distribution PD (K _i, t) between end of the call K _i using the parameters and the call K _i
And a fourth step of calculating a cell loss rate PB (K _i ) between the ends. Further, the waiting time distribution PD (K _i , t) obtained in the second step and the cell loss rate PB (K
_i ) a fifth step of determining whether or not the predetermined required quality is satisfied; and determining whether there is a path connectable to the call K _i. A sixth step of executing the first step and the second step when it is determined, and the call K _i when it is determined that the connectable route does not exist in the fourth step. The connection is rejected, and the waiting time distribution PD (K _i , t) and the cell loss rate PB (K
_A seventh step of connecting the call K _i when it is determined that _i ) satisfies the predetermined required quality may be provided.

[0006]

DESCRIPTION OF THE PREFERRED EMBODIMENTS ATMs studied in the present invention
FIG. 1 shows a network model. ATM network consists of five transit exchange _{(Node N 1 ~Node N 5)} , all of which are connected in series. Assuming that the cell switching function of each ATM switch is ideal (no delay, no loss) and the cell multiplexing function can be modeled by one buffer (queue), the ATM network of FIG. The model can be rewritten as a serial queue model as shown in FIG. Here, the node N _i arrived, the node N _{j (j>} i) All cell single cell source which leave the G
_ij , and its occurrence rate is R (i, j) [cells
/ Sec]. Assume that each cell source is independent of each other.

(Case where Cell Arrival Process is Poisson) (1) Simulation Model First, a case where the cell generation process of the cell source G _ij is all Poisson will be considered. The cell removal speed C _p of each processor is C _p = 100 [cells / sec], and the buffer capacities B are all B = 80 [cells]. Assuming that the sum of the cell arrival rates in the buffer X _i is R _i , the load factor ρ _i = R _i / C _p of the buffer X _i is always ρ = 0.95 regardless of i according to the procedure of FIG. Determine the cell generation rate for each cell source. Simulation time is 2 × 1 each time
It is set to 0 ⁵ [sec], and the average value of the judgment values obtained by the seven simulations is used as the measurement result. The simulation is performed by changing the load pattern of G _ij in 300 ways, and the cell loss rate in each buffer and the delay time distribution between the ends of G ₁₆ are measured.

(2) Simulation Results The average cell rate of the cell traffic input from the buffer X _i-1 to the buffer X _i is represented as F _i [cells / sec]. F _i
Is determined by the following equation.

[0009]

(Equation 1) Here T _i [cells / sec] represents the amount of traffic to leave the network at node N _i.

FIG. 4 shows the relationship between F _i and the cell loss rate at each of the nodes N ₂ to N ₅ . The dotted line in FIG. 4 indicates the cell loss rate in the M / D / 1 / K queue having a load factor of 0.95 and a buffer length of 80 [cells]. From FIG. 4, it can be seen that the cell loss rate in the M / D / 1 / K queuing model is almost the upper limit of the cell loss rate at each node. Also, since the cell loss rate tends to decrease as Fi increases, and since this characteristic is the same at every node, by modeling each node with an M / D / 1 / K queue, It is considered that the upper limit of the cell loss rate of each node can be evaluated.

Next, attention is paid to the cell delay time. The probability that the cell delay time between ends is equal to or longer than T [sec] is represented by P _t (T)
It expresses. FIG. 5 shows the sum S of F _i .

(Equation 2) P _t (1.8) with respect to. The dotted line indicates the probability that the delay time when each buffer is regarded as an independent M / D / 1 queue is 1.8 [sec] or more. From this figure, it can be seen that _Pt (1.8) tends to decrease as S increases. FIG. 6 shows an example of the cell delay distribution in four different S. The dotted line in the figure shows the cell delay time distribution between the ends when each buffer is regarded as an independent M / D / 1 queue. From the figure,
As S is larger, the peak of the cell delay distribution shifts to the left, while the tail of the cell delay distribution is M / D /
It can be seen that the delay distribution approaches the delay distribution of the multi-stage model with one queue.

(Case where Cell Arrival Process is Bernoulli) (1) Simulation Model In an actual ATM network, the number of input / output interfaces per exchange may be relatively small (less than 10). In such a case, the influence of the number of simultaneous arrivals of cells limited by the number of ports cannot be ignored. Therefore, the Bernoulli model is more suitable than the Poisson model as a model when cells arrive randomly. Here, the traffic characteristics when all the cell sources in the model of FIG. 2 are replaced by Bernoulli cell sources will be investigated. Cell interval time of each cell source is fixed to all 1 / C _p, the load pattern in the case as well as G _ij of the Poisson model 30
The simulation was performed with 0 changes.

[0013] (2) Simulation results First, attention is focused on the cell delay characteristic at the node N _1. The standard deviation σ ₁ of the cell generation rate of each cell source connected to the node N ₁ is defined by the following equation.

[0014]

(Equation 3) It is known that the distribution of the time required for Bernoulli traffic to pass through the buffer approaches an exponential distribution.
Therefore, a coefficient r ₀ when the distribution of the cell delay time t when passing through the node N ₁ is approximated by an exponential function C ₀ exp (−r ₀ t) is defined as a slope of the cell delay distribution.

FIG. 7 shows the slope r of the cell delay distribution with respect to σ ₁ .
₀ is shown. The crosses in the figure indicate Geo when the cell arrival probability q per time slot and the number D of time slots per service time (1 / C _p ) satisfy qD = ρ and D is changed from 2 to 5. / D / 1 queuing model [(Reference 3: Annie Gravey. Jean-Raymond Louvion
and Pierre Boyer, "On the Geo / D / 1 and Geo / D / 1 / n Q
ueues ", PerformanceEvaluation, vol.11, pp.117-125,
1990.)]. Note that σ = σ _G when the Geo / D / 1 queuing model is used is obtained by the following equation.

[0016]

(Equation 4) Here, N represents the number of input ports of the exchange of interest. From FIG. 7, no matter how varied the combination of cell incidence of each cell sources, the slope of the cell delay distribution in the node N ₁ is seen may depend on sigma ₁ value. That is, the cell delay distribution at node N ₁ may be characterized by a standard deviation sigma ₁ cell source that is input to the node N _1. Moreover, it increases monotonically with respect to the inclination of the distribution sigma _1, since the maximum value in the case of sigma ₁ = 0 load are equal to each other of the input traffic at each node,
It is considered that the upper limit of the cell delay distribution when the cell arrival process of input traffic is Bernoulli can be evaluated as follows using the Geo / D / 1 queuing model.

Step 1: Input a cell into the buffer of interest
Standard deviation σ of the cell generation rate of the cell source ₁Ask for. Step 2: σ₁Maximum σ not exceeding_GFind D that gives
This is D_MAnd Step 3: q = ρ / D_M, D = D_MGeo / D / 1
Find the waiting time distribution of the queue. This distribution is
Is the upper limit of the delay time distribution of traffic passing through
You.

FIG. 8 shows the cell loss ratio with respect to the standard deviation σ ₁ of the input traffic at each node. Note that σ _{i at} i> 1 is obtained by the following equation.

(Equation 5)

Here, F _i is obtained from (1) and (2). The mark x in FIG. 8 indicates that the load factor and the buffer length are equal.
9 shows the cell loss rate in the eo / D / 1 / K queue. From FIG. 8, it can be seen that the cell loss rate tends to decrease as σ _i increases, but the deviation from the cell loss rate of the Geo / D / 1 / K model increases toward downstream nodes, and σ _i The correlation between _i and the cell loss rate is no longer clear. Therefore, the downstream node cannot estimate the cell loss rate based on the value of the standard deviation of the input traffic, but the upper limit of the cell loss rate is set to Ge of D = 5.
It can be evaluated by the cell loss rate of the o / D / 1 / K model. However, it should be noted that the results shown in the present application cannot expect high accuracy for a cell loss rate of 10 ⁻⁶ or less from the upper limit of the simulation time.

FIG.₁+ Σ_TwoCell delay between end and end
Probability P that the interval is 1.5 [sec] or more _t(1.5) Relationship
It is shown. The dotted line indicates that each buffer has an independent D =
5 when compared to the Geo / D / 1 queue of 5
_t(1.5) is shown. From FIG. 9, σ₁+ Σ_Two
The larger the value of P_t(1.5) tends to be smaller
You can see that FIG. 10 shows four different σs.₁+ Σ_TwoTo
2 shows an example of a cell delay distribution in the present embodiment. In FIG.
Dotted line indicates Geo of D = 5 with equal load factor and buffer length.
/ D / 1 cell delay time when passing through 5 queues
Shows cloth. From FIG. 10, the tail of the cell delay distribution is Ge
It can be seen that it approaches the distribution of the series model of o / D / 1
You.

[0021]

An embodiment of the present invention will be described. FIG. 11 shows an example of the configuration of a switch of each node used in the present invention. 11 is a line terminating device on the incoming line, 12 is a cell switch, 13 is a buffer for multiplexing cells, and 14 is a cell extraction processor. , 15 is a line terminating device on the outgoing line side,
16 is a route selection function, 17 is an exchange function control unit, 18 is a buffer control unit, and 19 is a call admission control unit that executes the main operation of the method of the present invention.

The operation of the method of the present invention will be described below with reference to FIGS. Figure 12 shows the outline of the operation of the present invention a method for call K _i, the call admission control is started (100), whether there is available connection path in the route selection function 16 determines Is performed (10
1). This embodiment (101) is not indispensable in the present invention, since an embodiment may be adopted in which the route has already been determined or this determination is not made. Next, a parameter for calculation described later is obtained by using information from the exchange side (12, 17), and is notified to an admission control function in the call admission control unit 19 (102). The call admission control function obtains a waiting time distribution and a cell loss rate using the calculation parameters (103). The above processes 102 and 103 are the main operations of the present invention. As an additional process, call K _i is determined whether or not to satisfy the predetermined required quality (104). When the determination processing 101 is yes, the processing 102, 1
Running 03,104, when the determination 104 is yes, connect the call K _i, the determination processing 101 is at the no rejects the call connection K _i, and ends the call admission control (10
6).

[0023] FIG. 13 shows a specific example of the route selection operation 101 calls K _i, the number n of the exchange through which the call K _i after the start of the operation is 1 (101-1). Next, call _Ki
J the ID number N _s of the exchange which has received the connection request
It is substituted for (K _i , n) (101-2). The exchange J
(K _i, n) is determined whether accommodating the receiving terminal of the call K _i (101-3). This judgment processing 101-3
There when no, and n = n + 1, the process of substituting the ID number of the exchange call K _i is then used to _{J (K i, n) (} 101-4), performs again the process 101-3. When the process 101-3 is yes, the number of nodes through which the M _i = n (101-5), the process ends 101.

FIG. 14 shows a specific example of the process 102. After initiating processing, in the processing 102-1 obtains the ID number X of the buffer used is (1) call K _i,
This is substituted for b (K _i , N _i ). (2) Capacity B (x) of buffer X and take-out speed C
Find (x). (3) A call set GC (x) using the buffer X is obtained. (4) The number NP (x) of input ports used by calls belonging to GC (x) is obtained. (5) Find the sum R (x) of the average cell rates of calls using buffer X. Next, in process 102-1
Obtained calculation parameters: B (x), C (x), N
Notify P (x) and R (x) to the call admission control unit.

FIG. 15 shows a specific example of the process 103. After the processing is started, a queue model is selected (10
3-1). Next, the number n of exchanges through which the call _Ki passes is set to 1
(103-2), the waiting time distribution Pd (K _i , n, t) and the cell loss rate Pb in the n (= 1) th switch
(K _i , n) is obtained (103-3). Then, until the number n of nodes through which the call K _i is equal to M _i (103
-4), the process 103-3 is repeatedly performed, and the process 103-3 is performed.
When -4 becomes yes, in the process 103-5, (1) the end-to-end wait time distribution PD (K _i , t) is obtained. (2) Find the cell loss rate PB (K _i ) between the ends.

FIG. 16 shows a specific example of the queue model selection process 103-1 in FIG. After initiating processing, the maximum value maxNP number of input ports that call K _i for using the buffer X is passing max [NP (b
(K _i , J (K _i , n))].
1). a n = 1~M _i. Next, it is determined whether the cell arrival is suitable for being treated as a continuous time model (103).
-1-2), if yes, the cell arrival process is modeled as a Poisson process (103-1-3). Processing 10
If 1-1-2 is no, the cell arrival process is modeled as a Bernoulli process (103-1-4).

FIG. 17 shows the processing 103-3 in FIG.
(A) of FIG. After processing starts,
It is determined whether or not the cell arrival process is a Bernoulli process (103-3-1). The determination process 103-3-2 is ye
s, the maximum number of arrivals per service slot D
= MaxNP, 1 service slot time = 1 / C
(X), load factor ρ = R (x) / C (x) Geo / D /
From the one-queue model, the waiting time distribution Pd (K _i , n,
t) is obtained (103-3-2). Further, the maximum number of arrivals per service slot D = maxNP, one service slot time = 1 / C (x), load factor ρ = R (x) /
The cell loss rate Pb (K _i , t) is obtained from the Geo / D / 1 queue model of C (x) and buffer capacity B (x) (103-3-3).

FIG. 18 shows the process 103-3 in FIG.
(B) of FIG. After the start of the process, the processes 103-3-1 and 103-3-2 in FIG. 17 are performed. When processing 103-3-2 is no, one service slot time = 1 / C (x), load factor ρ = R (x) /
From the M / D / 1 model of C (x), the waiting time distribution Pd
(K _i , n, t) is obtained (103-3-5). Further, one service slot time = 1 / C (x), load factor ρ
= R (x) / C (x), M / D of buffer capacity B (x)
The cell loss rate Pb (K _i , n) is obtained from the / 1 / K model (103-3-6).

FIG. 19 shows the process 103-5 in FIG.
Shows a specific example of part (1) of the above, and the end-to-end waiting time distribution PD (K _i , t) is

PD (K _i , t) = Pd (K _i , 1, t) * Pdt (K _i , 2, t) *... * Pdt (K _i , M _i , t) (7) (103-5 (1) -1). Here, * is an operator representing a convolution operation.

FIG. 20 shows the process 103-5 in FIG.
Shows a specific example of a part (2) of the above. The end-to-end cell loss rate PB (K _i )

(Equation 7) Is obtained. (103-5 (2) -1).

As a specific example of the required quality in the processing 104 in FIG. 12, for example, the cell loss rate PB (K _i ) ≦ CLR (K
_i ) Maximum value of waiting time between ends ≦ maxCTD (K
_i ) The variation range of the waiting time from end to end ≦ CDV (K _i ). Each value on the right side is a predetermined limit value of the value on the left side of the corresponding equation.

[0032]

As described in detail above, a plurality of ATs
To investigate the characteristics of the traffic passing through the M switch,
According to the present invention, simulations were performed by changing the traffic route and load distribution in the network in various ways. As a result, if all the input traffic is Poisson,
Regardless of the traffic route and load distribution, the model considers each node as an independent M / D / 1 and M / D / 1 / K in a serial queue, and the cell delay time distribution and the upper limit of the cell loss rate of the traffic of interest are modeled. It turns out that it can be evaluated. When the number of input ports of the ATM switch is small, the delay time distribution and the upper limit of cell loss are stricter by a serial queuing model in which each node is regarded as an independent Geo / D / 1 and Geo / D / 1 / K. It was found that it could be evaluated. Therefore, the present invention provides a
Since the evaluation of the traffic characteristics of the ATM network including the M exchange can be appropriately executed, and the result can be used for line management and appropriate route setting, the practical effect is extremely large.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a model of an ATM network to which the present invention is applied.

FIG. 2 is a block diagram for explaining the principle of the present invention.

FIG. 3 is a flowchart showing a procedure for setting a cell generation rate used in a simulation for confirming the effectiveness of the background art of the method of the present invention.

FIG. 4 is a characteristic diagram showing a measurement example of a cell loss ratio for each node measured according to the present invention.

FIG. 5 is a characteristic diagram of a cell delay time distribution measured according to the present invention.

FIG. 6 is a characteristic diagram of an end-to-end delay time distribution measured according to the present invention.

FIG. 7 is a characteristic diagram showing a relationship between a standard deviation of a cell arrival rate measured according to the present invention and a slope of a cell delay time distribution.

FIG. 8 is a characteristic diagram showing a measurement example of a cell loss ratio with respect to a standard deviation of input traffic at each node measured according to the present invention.

FIG. 9 is a diagram illustrating an example of measurement of an end-to-end cell delay distribution according to the present invention.

FIG. 10 is another measurement example diagram of the end-to-end cell delay distribution according to the present invention.

FIG. 11 is a block diagram showing an embodiment of the present invention.

FIG. 12 is a flowchart for explaining the operation principle of the method of the present invention.

FIG. 13 is a flowchart showing a specific example of a process 101 in FIG. 12;

FIG. 14 is a flowchart showing a specific example of a process 102 in FIG. 12;

FIG. 15 is a flowchart showing a specific example of a process 103 in FIG. 12;

FIG. 16 is a flowchart showing a specific example of a process 103-1 in FIG. 15;

FIG. 17 is a flowchart showing a specific example of a process 103-3 in FIG. 15;

FIG. 18 is a flowchart for explaining another specific example of the processing 103-3 in FIG. 15;

FIG. 19 is a flowchart for explaining a specific example of the processing 103-5 (1) in FIG. 12;

FIG. 20 is a flowchart illustrating a specific example of a process 103-5 (2) in FIG. 12;

[EXPLANATION OF SYMBOLS] 11 Line terminator 12 cell exchange 13 buffer 14 cell extraction processor 15 line terminating equipment 16 of the route selection function 17 exchange function controller 18 buffer control unit 19 the call admission control unit K _i call ID M _i call K _i There total number J (K _i, n) of nodes through _{ID b (K i, n i} ) of the call K _i is the n-th ID n _i node of the exchange through which (1 to M _i) (exchange) call K _i ID of the buffer used by the switch N _i X ID of the buffer B (x) Capacity of the buffer X GC (x) Set of calls using the buffer X NP (x) Input through which the call using the buffer X is passing number Pd port (K _i, n, t) cells time t wait probability density (latency distribution) in exchange through the n-th call _{K i Pb (K i, n} ) cell n-th call K _i To the exchange via There are probability PD (K _i, t) is lost probability density (latency distribution) wait time t between the end of the cell call K _i PB (K _i) loss between end cell of the call K _i (cell loss ratio R (x) Sum of average cell arrival rates of traffic input to buffer X C (x) Cell take-out speed of buffer X Am (x) Design value of buffer X load factor (upper limit)

Claims (7)

[Claims] For the evaluation of 1. A cell traffic characteristic through a plurality of ATM exchanges, in desired the ATM switch of the plurality of ATM exchanges through which the call K _i to be evaluated, the cells of the cell traffic A first step for determining a type of arrival process; a second step of selecting a queue suitable for determining the cell traffic characteristics according to the determined type of cell arrival process; a third step of using the queue determine the parameters required to determine the cell traffic characteristics, latency distribution PD (K _i, t) between end of the call K _i using the parameters and the call K _a fourth step of determining a cell loss rate PB (K _i , n) between the end of _i and the cell traffic through a plurality of ATM exchanges. Evaluation method 2. The method according to claim 1, wherein the fourth step comprises: selecting the plurality of ATs using the selected queuing model.
Latency distribution Pd (K _i , n, t) and cell loss rate Pd (K _i , n) in the n-th switch among the M switches
4-1 for all of the plurality of ATM exchanges through which the call K _i passes to obtain the end-to-end latency distribution PD (K _i , Step 4-2 for calculating the cell loss rate PB (K _i ) between t) and the end.
The plurality of ATs according to claim 1, comprising:
A method for evaluating cell traffic characteristics via an M switch. 3. In the second step, when it is determined that the cell arrival is suitable for being treated as a continuous time model, the cell arrival process is modeled as a Poisson process,
The plurality of A according to claim 1, wherein the cell arrival process is modeled as a Bernoulli process when it is determined that the cell arrival is suitable for being treated as a discrete model.
Evaluation method of cell traffic characteristics via TM exchange. Wherein when a cell arriving at the second step is modeled as a Poisson process, the parameters need to be determined in the third step, the call K _i in the relevant ATM switching system is used Of the buffer X of the buffer X, the cell removal rate C (x) of the buffer X, the number NP (x) of input ports through which calls using the buffer are passing, and the input to the buffer X. 2. The method for evaluating cell traffic characteristics via a plurality of ATM exchanges according to claim 1, wherein the sum is the sum R (x) of average cell arrival rates of traffic. 5. When the cell arrival is modeled as a Bernoulli process in the second step, the required parameters determined in the third step are the capacity B (x) of the buffer X of the ATM switch. The cell extraction rate C (x) of the buffer X, the number NP (x) of input ports through which calls using the buffer X pass, and the sum R of the average cell arrival rate of traffic input to the buffer X 2. The method for evaluating cell traffic characteristics via a plurality of ATM exchanges according to claim 1, wherein (x) and a maximum value maxNP of said parameter NP (x). 6. The maximum value ma of the parameter NP (x)
When the maximum arrival number D per service slot corresponding to xNP is 10 or more, it is determined that the cell arrival is suitable for being treated as a continuous time model.
4. The method according to claim 3, wherein the cell arrival number is determined to be suitable for handling as a discrete model when is less than 10. 7. The waiting time distribution PD (K _i , t) obtained in the second step and the cell loss rate PB
A fifth step of determining whether (K _i ) satisfies a predetermined required quality; and determining whether there is a path connectable to the call Ki _; A sixth step of executing the first step and the second step when it is determined that the call exists, and a call K when it is determined that the connectable path does not exist in the fourth step. refuse the connection of _i , the third
The to the Step latency distribution PD (K _i, t) to connect the call K _i when said cell loss probability PB (K _i) is determined to satisfy the required quality of the predetermined 7. The method for evaluating cell traffic characteristics via a plurality of ATM exchanges according to claim 1, further comprising: