JP2001044998A - Method and device for estimating communication quality - Google Patents

Abstract

PROBLEM TO BE SOLVED: To estimate the end to end communication quality by checking the goodness of fit of plural linear multiple regression models and inputting the management information base(MIB) information to an optimum fitting model to estimate the end-end quality. SOLUTION: The even background load are generated by a desktop server 1, etc., to the terminals 19a-19d respectively and at the same time the performance measurement traffic is flowed to the terminal 19a to measure the throughput. The terminals 19a-19d and a MIB collection terminal 5 collect the MIB data on an upstream router 9 of a transmitting router, the subscriber's routers 17a-17d and a high function subscriber's router 21 respectively. The experiment condition of every category is quantized since both experiment and MIB information are treated at a time, and an optimum variable is selected for setting the end-end performance. A most fitting combination of variables is selected to be used for estimating the end-end performance among the MIB information and quantized experiment conditions. The MIB information is inputted to an optimum model and the end to end quality is estimated.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a communication quality estimating method and apparatus for estimating and managing communication quality in a computer network.

[0002]

2. Description of the Related Art Quality that is meaningful to users of computer networks is end-to-end quality, such as end-to-end throughput, but end-to-end quality information is not readily available. What can be easily obtained is quality information in each NE (network component) in the network, for example, MIB (Mana) in an IP network.
gement Information Base (management information base) information and the like, it is desired to perform quality control of a computer network using such easily available quality information.

[0003]

As described above, since the end-to-end quality information is not easily available, the end-to-end quality is estimated by using the easily obtainable quality information at each NE in the network. While it is desirable to control the quality of a computer network, such a technique has not existed.

[0004] The present invention has been made in view of the above,
It is an object of the present invention to provide a communication quality estimation method and apparatus for estimating end-to-end communication quality in a computer network using quality information that can be easily obtained by network components.

[0005]

In order to achieve the above object, the present invention according to the first aspect of the present invention is directed to an information processing apparatus which can obtain MIB information and end-to-end information from various network components in a network under various network configurations and background traffic conditions. The quality of non-numerical data, such as network configuration and background traffic, is quantified by quantification theory. The gist is to examine the degree of fit of the linear multiple regression model, select the model that fits best as the optimal model, and input MIB information to the optimal model to estimate end-to-end quality.

According to the present invention, MIB information and end-to-end quality are measured from various network components under various network configuration and background traffic conditions, and data is obtained. Of the non-numerical data is quantified by quantification theory, the obtained numerical data is checked for the degree of fit of a plurality of linear multiple regression models, and the best-fitting model is selected as the optimum model, In order to estimate the end-to-end quality by inputting the MIB information to the NE, it is possible to accurately estimate the end-to-end quality from the easily obtainable MIB information in each NE.

The present invention according to claim 2 is based on claim 1.
In the invention described above, the gist is that the process of acquiring the data uses an experiment design method that optimizes the number of experiments.

According to the second aspect of the present invention, data can be efficiently obtained with a small number of experiments because the number of experiments is optimized by using an experiment design method.

Further, according to the present invention, there is provided an input means for inputting MIB information and data obtained by measuring end-to-end quality from respective network components in a network under various network configurations and background traffic conditions. And quantification means for quantifying non-numeric data such as network configuration and background traffic in the input data by quantification theory, and a plurality of linear multiple regression models for the obtained quantified data. Model selection means for examining the degree of conformity of the model and selecting the best-fitting model as the optimal model.
An estimating means for estimating end-to-end quality by inputting information.

According to the present invention, MIB information and data obtained by measuring end-to-end quality are input from each network component under various network configurations and background traffic conditions. Of these, non-numerical data is quantified by quantification theory, the obtained numerical data is checked for the fitness of a plurality of linear multiple regression models, and the best-fitting model is selected as the optimal model, and the optimal model is selected. Since the MIB information is input and the end-to-end quality is estimated, the end-to-end quality can be accurately estimated from the easily obtainable MIB information in each NE.

The present invention according to claim 4 obtains data by measuring MIB information and end-to-end quality from each network component in the network under various network configurations and background traffic conditions. Non-numerical data such as network configuration and background traffic is quantified by quantification theory, and the obtained numerical data is examined for the degree of fit of multiple multivariate polynomial regression models. The gist is to select a suitable model as an optimal model, input MIB information to the optimal model, and estimate end-to-end quality.

According to the present invention, MIB information and end-to-end quality are measured from various network components under various network configuration and background traffic conditions, and data is obtained. Of the non-numerical data is quantified by quantification theory, the obtained numerical data is examined for the degree of fit of a plurality of multivariate polynomial regression models, and the best-fitting model is selected as the optimal model. Since the end-to-end quality is estimated by inputting the MIB information to the model, even if the end-to-end quality cannot be linearly approximated to the easily obtainable MIB information at each NE, the end-to-end quality can be accurately determined from the MIB information. Can be estimated.

The present invention according to claim 5 provides the present invention according to claim 4.
In the invention described above, the gist is that the process of acquiring the data uses an experiment design method that optimizes the number of experiments.

According to the fifth aspect of the present invention, data can be efficiently obtained with a small number of experiments because the number of experiments is optimized by using an experiment design method.

Further, according to the present invention, there is provided an input means for inputting MIB information and data obtained by measuring end-to-end quality from various network components in a network under various network configurations and background traffic conditions. And quantification means for quantifying non-numerical data such as network configuration and background traffic in the input data by quantification theory, and a plurality of multivariate polynomial regressions on the obtained quantified data. A model selecting means for examining the degree of fit of the model and selecting the best-fitting model as the optimal model.
An estimator for inputting IB information and estimating end-to-end quality is provided.

According to the sixth aspect of the present invention, MIB information and data obtained by measuring end-to-end quality are input from each network component under various network configurations and background traffic conditions. Of these, non-numerical data is quantified by quantification theory, the obtained numerical data is examined for the fitness of a plurality of multivariate polynomial regression models, and the best-fitting model is selected as the optimal model. To estimate the end-to-end quality by inputting the MIB information to the end-to-end quality accurately even if the end-to-end quality cannot be linearly approximated to the easily obtainable MIB information at each NE. can do.

[0017]

Embodiments of the present invention will be described below with reference to the drawings. FIGS. 1 to 3 are diagrams illustrating a method for estimating a communication quality according to an embodiment of the present invention.
It is a block diagram which shows the structure of each type of network NW1, NW2, and NW3. 1 to 3, reference numeral 1 denotes a desktop server (DT), 3 denotes a notebook server (NB), and 5 denotes an MIB collection terminal (MI).
B), 7 are hubs, 9 is an upstream router, 11 is a local line terminal (SLT), 13 is a star coupler (SC), and 15a to 15d are subscriber line terminals (a). ONU), 17a to 17d are subscriber premises routers (m
R), 19a to 19d are terminals, 21 is a high function subscriber premises router (Cisco), and 23 is a hub.

The three types of networks N shown in FIGS.
1 and 3, the networks NW1 and NW3 shown in FIG. 1 and FIG.
T) 11 and each terminal 19 at a star coupler (SC) 13 between the subscriber line termination unit (ONU) 15 and the network NW2 shown in FIG. It is branched at a hub (Hub) 23 connected to a high-function subscriber premises router (Cisco) 21.

In the network used in this experiment, 10 subscriber units (ONUs) 15
A bandwidth of 0 Kbps is guaranteed. Therefore, in the networks NW1 and NW3 shown in FIGS. 1 and 3, a bandwidth of 100 Kbps is guaranteed for each terminal.

In the networks NW1 and NW3 shown in FIG. 1 and FIG. 3, the router connected to the terminal 19a is the subscriber premises router (mR) 17a in FIG. (Cisco) 21, but the advanced subscriber premises router (Cisco) 21 used in the network NW3 is the subscriber premises router (mR) 1 used in the network NW1.
More packets can be processed as compared to 7a. Therefore, if the subscriber premises router (mR) 17a is a bottleneck in the network NW1, there will be a difference in network performance.

In FIG. 1 to FIG.
That is, the desktop server 1, the notebook server 3, the MIB collection terminal 5, the upstream router 9, the subscriber premises routers (mR) 17a to 17d, and the terminals 19a to 19
d. The high-function subscriber premises router (Cisco) 21 can communicate with each other using the Internet Protocol (IP).

Each of the upstream router 9 and the customer premises router (mR) 17 has a management information base (MI
B), and its element includes the number of octets of data transmitted and received via each interface. In addition, the value can be referred to from another device using SNMP.

Each of the terminals 19a to 19d and the MIB collection terminal 5 is a subscriber premises router (mR) 17a to 1
7d, each MIB object of the upstream router 9 can be learned using SNMP. In addition, the load on the network and the CPU of each device at the time of collection is sufficiently smaller than that of other communication, and
The effect of collecting MIB information by using on other communications can be ignored.

The time of each of the terminals 19a to 19d, the desktop type server 1, the notebook type server 3, and the MIB collection terminal 5 is determined by the Network Time Protocol (NTP).
Are synchronized using

The amount of data (number of octets) of packets transmitted and received through each interface of the router is collected every second using SNMP, and the difference is obtained. The amount of data transmitted and received through the interface can be determined.

Software DBS for generating traffic is installed in the terminals 19a to 19d, the desktop server 1, and the notebook server 3, so that any traffic can be generated between the terminals. Further, with respect to the generated traffic, performance information such as throughput and delay can be recorded at the receiving end.

Using the network configured as described above, a method of selecting a management data item most suitable for estimating end-to-end performance, a method of estimating end-to-end performance, and a management goal of end-to-end performance given in advance The procedure for obtaining the management threshold from the values through experiments will be described.

First, as a performance measure to be estimated by the communication quality estimating method of the present embodiment, that is, an end-to-end performance, a traffic for performance measurement under the conditions described in Table 1 shown in FIG. Let us take up the throughput on the receiving side when flowing to the respective networks NW1, NW2 and NW3 shown in FIG.

The data used for estimating the end-to-end performance includes, as information that can be collected from the MIB information, the transmission traffic volume of the upstream router 9 and the customer premises router (mR) 17 or the high-function subscriber premises router ( Cisc
o) The received traffic volume of the downstream router consisting of 21 is used.

Further, the following experimental conditions are set.

(1) The network N shown in FIGS.
Configuration of W1, NW2, NW3 (2) Protocol ratio of load traffic (TCP: UDP = 1: 0, 1: 1, 0: 1) (3) Packet size of load traffic (64, 300, 1500 bytes) (4) ) Number of load flows (2, 4, 8) (5) Type of load generation server (Desktop server: Notebook server = 1:
(0, 1: 1, 0: 1) (6) Experiment time or number of experiments (first, second, third) Note that the assumed total background load (4.0, 3.6, 3,.
Experimental conditions of (2,2.8,2.4,2.0 Mbps) were not used for estimation.

Next, as a background load generation method, traffic serving as a background load is generated according to the experimental conditions set as described above. That is, traffic is generated such that the assumed total background load, the packet size, the protocol ratio, and the load generated by each server meet the above experimental conditions, and the packet generation frequency is evenly spaced.

Then, end-to-end performance measurement and MIB data collection are performed while generating traffic under the conditions selected as described above. That is, the above-described background load is evenly generated from the desktop server (DT) 1 or the notebook server (NB) 3 located upstream of the network shown in FIGS. 1 to 3 toward the terminals 19a to 19d. Meanwhile, the traffic for performance measurement shown in Table 1 shown in FIG. 4 is simultaneously sent from the desktop server 1, which is the upstream server, to the terminal 19a, and the terminal 19a, which is the receiving terminal, measures the throughput. At the same time, the upstream router 9 which is the transmitting router and the subscriber premises routers (mR) 17a to 17d which are the receiving routers.
And M in the advanced customer premises router (Cisco) 21
The IB data is collected by the directly connected terminals 19a to 19d and the MIB collection terminal 5.

In the above-described processing, if the performance is measured for all combinations, the correspondence between the data from the network component (NE) and the end-to-end performance can be grasped more accurately, but it takes a lot of time and effort. Therefore, in order to grasp the correspondence with as few experiments as possible while considering as many factors as possible, the experiment design method L18 (Taichiro Ueda, Taguchi Method Complete Understanding Manual, Japan Business Report, 1998, Oc
t) was used.

As a result, the raw data obtained by the experiment is as shown in Table 2 shown in FIG.

Next, quantification processing of non-numerical data (categorical data) will be described. That is, in order to handle the experimental conditions and the MIB information simultaneously, the experimental conditions classified into categories are quantified. Specifically, since the above-described experimental conditions are data (categorical data) representing types without ordering, they cannot be used for regression analysis as they are. Therefore, the experimental conditions are converted into numerical data by quantification method 1. Table 3 shown in FIG. 6 is a numerical representation of the experimental conditions. However, U (assumed load amount)
Is the assumed load, but since MIB data is collected for the load, this MIB data is used and is not included in the converted table.

Here, regarding the method of creating Table 3, "proto
(Protocol ratio) ”as an example.
(See Statistical Dictionary, Takeuchi et al., Pp. 368-370).

First, TCP, T &
An arbitrary reference element is selected from U and UDP.
In this example, UDP is selected.

Then, one categorical variable proto
For “UDP → TCP” and “UDP → T & U”
If two variables are prepared and the measurement condition is TCP for each measurement, “UDP → TCP” is set to 1 and the measurement condition is T &
In the case of U, “UDP → T & U” is set to 1. The other columns are set to 0.

The meaning of the coefficient for each variable created here is “difference in estimated value when proto is changed from UDP to TCP” in the case of UDP → TCP.

Next, the optimum variables for estimating the end-to-end performance are selected. From the MIB information and the quantified experimental conditions (hereinafter referred to as candidates for explanatory variables), a combination most suitable as a variable used for predicting end-to-end performance is selected. In other words, among the data used for estimation, some items are selected as explanatory variables, and when a linear multiple regression model is considered with the performance scale to be estimated as the objective variable, the cases where each explanatory variable is included in the model There are two possible cases.

When considering all variables, a model of a combination of 2 raised to the "number of variables" can be considered. From these models, the optimal one is selected, and a scale as described below can be used as a scale representing the fitness at that time. In implementation, it takes time to check all models. (1) An initial model is arbitrarily determined.

(2) If each variable is included in the current model, consider a model excluding the variable. If not included in the current model, consider a model to which the variable is added. Consider a model with only the number of variables.

(3) Regarding the model considered as described above, an optimal model is considered.

(4) If the optimal model considered in this way is the same as the current model, the model is set as a (local) optimal model. Otherwise, the model considered in (3) is set as the current model, and the process returns to (2).

With the above procedure, a (local) optimal model is obtained. Since the fitness is monotonically decreasing and the number of possible models is finite, the fitness can be reduced by two steps at most.
The process is terminated by repeating the number of variables to the power of “power” (variable decrementing method in S).

As a scale indicating the degree of fit of the above-described model, Akaike Information Criterion (AIC) (Information Statistics, Sakamoto et al.,
pp.138-140, see equation (8.33)). That is, for the model of interest, the coefficients for each variable are determined using multiple regression analysis (for the coefficient determination method, see Information Statistics, Sakamoto et al., Pp.138-139, equation (8.30)). Once the coefficient is determined,
An Akaike information criterion (AIC) can be calculated for the model. A model with a smaller AIC is more suitable. That is, it can be said that the set of variables included in the model is suitable for end-to-end performance prediction. The regression equation at this time is a prediction equation for end-to-end performance.

Table 4 shown in FIG. 7 shows the process of selecting the objective variable. In Table 4, the column of “step” indicates how many times the operation of taking in and out of the variable has been performed. Next, each row will be described.

In step 0 in Table 4, a model containing no explanatory variables is considered. The AIC at that time is described in the “current AIC” column. The right column shows the AI when each variable was added to the model as an explanatory variable.
C is represented. The smallest value among the values entered in the step 0 line is -201.10 in the "loss rate" column.
Therefore, in the next step, a model using “loss rate” as an explanatory variable will be considered.

In the next step 1, a model in which “loss rate” is used as an explanatory variable and the AIC when each variable is added to the model are shown. However, the column indicated by parentheses () indicates the AIC when the corresponding variable is excluded from the model. The smallest value among the numerical values entered in the step 1 line is -201.40 in the column "number of flows 8 → 2".

In step 2, it is considered to increase or decrease the explanatory variables from a model that takes “loss rate” and “the number of flows 8 → 2” as explanatory variables. The smallest value among the numerical values entered in the step 2 line is the “current AIC” column. This indicates that there is no more suitable model among those obtained by increasing or decreasing one variable from the current model. Therefore, the current model is set as a local optimal solution.

The model obtained here is as follows: y = (UDP throughput), x ₁ = (loss rate), x ₂ = (quantification variable meaning changing the number of flows from 8 to 2)

Y = (5.2 × 10 ⁻³ ) − (2.2 × 10 ⁻³ )
It can be written as x _1- (3.3 × 10 ⁻⁴ ) x ₂ .

Table 5 shows the statistics related to the regression coefficients.
Shown in Here, a ₀ is a constant term, and a ₁ and a ₂ are each x
_1, represents the regression coefficient of x _2.

Next, the result is considered and the validity of the selected variable is verified. Of the experimental conditions,
If any of the explanatory variables are used, their validity is verified using a test. With the following hypothesis,
The F test was attempted at a significance level of 5%.

There is no correlation between the null hypothesis, that is, the UDP throughput and the number of flows.

As a result of the analysis of variance, the upper probability of the F value is 0.65 and cannot be rejected. Therefore, it cannot be said that there is a correlation from the measured data.

Therefore, the result of regression of y only by x ₁ is again shown in Table 6 shown in FIG.

## EQU2 ## Model as y = (5.1 × 10 ⁻³ ) − (2.1 × 10 ⁻² ) × ₁ .

Next, a graph is drawn, and a management threshold value for a loss rate used for management is derived from a target end-to-end performance. The regression line of the model obtained as described above and the 95% confidence interval of the estimated value of the UDP throughput are shown in FIG.
0 is shown.

As an example of the management target, the management threshold value for the management target of maintaining the throughput at 0.004 Mbps on average when the measurement traffic flows as used in this experiment is shown in FIG. The average loss rate of 5% at the neck may be used as the threshold.

Next, a communication quality estimation method according to another embodiment of the present invention will be described.

In the embodiment shown in FIGS. 1 to 10 described above, the end-to-end quality information which cannot be easily obtained is obtained by using the MIB information which is the quality information which can be easily obtained by the NE in the network which can be easily obtained. The communication quality estimation method to be obtained has been described. In this method, good accuracy can be obtained when the end-to-end quality is close to linear with respect to the quality information obtained from the NE, but when linear approximation cannot be performed, the estimation accuracy may not be very good.

Therefore, in the communication quality estimation method of this embodiment, a multivariate polynomial regression model is used instead of the linear multiple regression model, whereby the end-to-end quality is linear with respect to the MIB information in each NE which can be easily obtained. Even if the approximation cannot be made, the end-to-end quality can be accurately estimated from the MIB information.

For further details, refer to FIGS.
Will be explained. In a network as shown in FIG.
And two lines that are likely to be bottlenecks in performance
Yes (bottleneck lines 115 and 121). like this
In the network, host 111 to host 125
When the round trip time y when a packet is reciprocated to
The case will be described. That is, the measured round trip time
Means that a packet for measurement is transmitted from the host 111 to the router 11
3, 117, 119, 123 to the host 125
From the host 125, the routers 123, 119,
Required to reach the host 111 through 117 and 113
Time. At the same time as the round trip time,
Usage rate x of the two-way line in the link lines 115 and 121
₁, X_Two, X _Three, X_FourWas also measured. Here, this line usage
Usage rate x₁, X_Two, X_Three, X_FourThe round trip time y from
The following equation is obtained when considering a model for

[0064]

(Equation 3) Here, a is a coefficient of a corresponding term, and n ₁ , n ₂ , n
_3, n ₄ is the order for _{x 1, x 2, x 3} , x 4, and _{a 10, a 20, a 30} , a 40 , respectively 0.

When the orders n ₁ , n ₂ , n ₃ , and n ₄ are determined, a is determined by the minimum AIC (Akaike information criterion), so that n ₁ , n ₂ , n ₃ , and n ₄ are plotted. 12 and FIG.
As shown in FIG. 3, it is determined by the stepwise method.

The model obtained as described above (n ₁ = 5,
The regression equation of (n ₂ = 0, n ₃ = 0, n ₄ = 1) is as follows.

[0067]

Equation 4] _{y = a 0 + a 11 x} 1 + a 12 x 1 2 + a 13 x 1 3 + a 14 x 1 4 + a 15 x 1 5 + a 41 x 4 ... (2) each regression coefficient and its statistics 14 Table 7 shown in FIG. This can be obtained by using two types of line utilization rates x ₁ and x ₄ among the samples used this time and substituting them into equation (2).
This indicates that the round-trip delay can be accurately estimated. Here, paying attention to x _4, t value upper probability of the highest order regression coefficient a ₄₁ is 0.093, not considered significant at significance of 5% t-test. On the other hand, the t-value upper probability of the regression coefficient a ₁₅ of the x ₁ highest order is 0.0014, indicating that x ₁ is very significant. Therefore, consider a polynomial model according to only x _1, was a regression equation was as shown in equation (3) shown below.

[0068]

Equation 5] _{y = a 0 + a 11 x} 1 + a 12 x 1 2 + a 13 x 1 3 + a 14 x 1 4 + a 15 x 1 5 ... (3) each regression coefficient and its statistic table shown in FIG. 15 8 When the measured values are superimposed on the regression equation and the 95% confidence interval, the results are as shown in FIG. From FIG. 16, it can be seen that the rapid increase in the delay due to the change in the usage rate can be appropriately modeled.

The model obtained by the communication quality estimating method of the embodiment shown in FIGS. 1 to 10 is given by the following equation (4), and the statistic related to the regression coefficient is as shown in Table 9 in FIG. . The AIC of this model is 1983
3.44.

Y = a ₀ + a ₁₁ x ₁ + a ₃₁ x ₃ (4) The AIC of the model obtained by the communication quality estimation method of the embodiment shown in FIGS. 1 to 10 is 19833.4.
The AIC of the model according to the present embodiment shown in FIG. 11 and subsequent figures is 19031.48 as compared with 4, which indicates that the present embodiment can better estimate the end-to-end quality.

[0071]

As described above, according to the present invention,
Under various network configurations and background traffic conditions, MIB information and end-to-end quality are measured from each network component to obtain data. Non-numerical data of this data is quantified by quantification theory. The degree of fit of a plurality of linear multiple regression models is examined with respect to the obtained numerical data, the model that fits best is selected as the optimal model, and MIB information is input to the optimal model to estimate the end-to-end quality. The end-to-end quality can be accurately estimated from the MIB information in each NE which can be easily obtained.

According to the present invention, MIB information and end-to-end quality are measured from various network components under various network configuration and background traffic conditions, and data is obtained. The data is quantified by quantification theory, the degree of fit of a plurality of multivariate polynomial regression models is checked with respect to the obtained quantified data, the best fit model is selected as the best model, and MIB information To estimate the end-to-end quality, so the MIBs at each NE that are readily available
Even when the end-to-end quality of the information cannot be linearly approximated, the end-to-end quality can be accurately estimated from the MIB information.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a network NW1 used to execute a communication quality estimation method according to an embodiment of the present invention.

FIG. 2 is a block diagram illustrating a configuration of a network NW2 used to execute a communication quality estimation method according to an embodiment of the present invention.

FIG. 3 is a block diagram illustrating a configuration of a network NW3 used to execute a communication quality estimation method according to an embodiment of the present invention.

FIG. 4 is a diagram showing Table 1 showing traffic for performance measurement in the embodiment of the present invention.

FIG. 5 is a diagram showing Table 2 showing raw data obtained in an experiment according to the embodiment of the present invention.

FIG. 6 is a diagram showing Table 3 showing data obtained by quantifying non-numerical data among the data shown in FIG. 5;

FIG. 7 is a diagram showing Table 4 showing a process of selecting an objective variable in the embodiment of the present invention.

FIG. 8 is a diagram showing Table 5 showing statistics regarding regression coefficients obtained by a variable increase / decrease method according to the embodiment of the present invention.

FIG. 9 is a diagram showing Table 6 showing statistics regarding regression coefficients of a final model in the embodiment of the present invention.

FIG. 10 is a diagram showing a regression line of a model and a 95% confidence interval of an estimated value of a UDP throughput in the embodiment of the present invention.

FIG. 11 is a diagram showing a network configuration for explaining a communication quality estimation method according to another embodiment of the present invention.

FIG. 12 shows the order n ₁ , n ₂ , and n in the model for estimating the line usage rate in the embodiment shown in FIG.
FIG. 9 is a diagram illustrating a part of a change of each step in a stepwise method for determining n ₃ and n ₄ .

FIG. 13 is a diagram showing the remaining part of the change of each step in the stepwise method following FIG. 12;

FIG. 14 is a diagram showing Table 7 showing coefficients of regression equations based on x ₁ and x ₄ and statistics thereof.

15 is a diagram showing a table 8 showing the coefficient and its statistical regression formula by x _1.

FIG. 16 is a graph showing a round-trip delay with respect to a line utilization rate (x ₁ ) by superimposing measured values on a regression equation and a 95% confidence interval.

17 is a diagram showing a table 9 shows the coefficients of the regression equation and its statistic by x _1.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Desktop type server 3 Notebook type server 5 MIB collection terminal 9 Upstream router 11 Station side subscriber line termination unit (SLT) 13 Star coupler (SC) 15 Subscriber side subscriber line termination unit (ONU) 17 Subscriber Home Router (mR) 19 Terminal 21 High Function Subscriber Home Router (Cisco)

Claims (6)

[Claims] The present invention obtains data by measuring MIB information and end-to-end quality from various network components in a network under various network configurations and background traffic conditions. , Non-numerical data such as background traffic is quantified by quantification theory, the obtained numerical data is examined for the fitness of multiple linear multiple regression models, and the best-fitting model is selected as the optimal model. And estimating end-to-end quality by inputting MIB information to the optimal model. 2. The communication quality estimating method according to claim 1, wherein the processing of obtaining the data uses an experiment design method that optimizes the number of experiments. 3. An input means for inputting MIB information and data obtained by measuring end-to-end quality from various network components in the network under various network configurations and background traffic conditions, and among the input data, Quantification method for quantification theory for non-numerical data such as network configuration, background traffic, and the likelihood of multiple linear multiple regression models are examined for the obtained quantified data to find the best fit. A communication quality estimating apparatus comprising: a model selecting unit that selects a model to be performed as an optimal model; and an estimating unit that inputs MIB information to the selected optimal model and estimates end-to-end quality. 4. Obtaining data by measuring MIB information and end-to-end qualities from various network components in the network under various network configurations and background traffic conditions. , Non-numerical data such as background traffic is quantified by quantification theory, and the obtained quantified data is examined for the fitness of multiple multivariate polynomial regression models. A communication quality estimating method characterized by selecting MIB information, inputting MIB information to the optimal model, and estimating end-to-end quality. 5. The communication quality estimating method according to claim 4, wherein the process of obtaining the data uses an experiment design method that optimizes the number of experiments. 6. Input means for inputting MIB information and data obtained by measuring end-to-end quality from respective network components in the network under various network configurations and background traffic conditions, and among the input data, Quantification means for non-numerical data such as network configuration, background traffic, and the like, and the degree of fit of multiple multivariate polynomial regression models to the obtained quantified data. A communication quality estimating apparatus comprising: a model selecting unit that selects a suitable model as an optimal model; and an estimating unit that inputs MIB information to the selected optimal model and estimates end-to-end quality.