JP3443531B2 - Communication simulation device - Google Patents

Description

Detailed Description of the Invention

[0001]

TECHNICAL FIELD The present invention is used for traffic quality evaluation of an information communication system. Mainly used for evaluation for traffic design and quality confirmation after system construction. The present invention branches and observes a packet passing through a path of interest in the middle of a switching system, actually detects the arrival time of the packet or a specific cell in the packet, and from the result, detects the packet or the packet. It is used as a device for estimating the arrival time of a large number of cells and the loss rate of the large number of cells included in.

[0002]

2. Description of the Related Art In an information communication system, quality such as processing delay and data loss varies depending on the amount of traffic. Therefore, in order to construct an economical and high quality information communication system, the buffer size, line speed, CPU power,
It is indispensable to provide an appropriate amount of buffers and the like based on the evaluation of how much traffic and what kind of quality the memory has, and the results.

To this end, two methods have mainly been used so far. The first method is simulation. Build a logic that simulates the target system on the computer. The data obtained from the traffic measurement items of the information communication system are statistically processed, traffic is generated by random numbers based on the result, and the quality generated for the traffic is evaluated.

The second method is a method of calculating a relational expression between a condition such as a traffic buffer size and quality obtained by a traffic theory or the like. As with the first method, the traffic condition is statistically processed from the data obtained from the traffic measurement items of the information communication system and given by the result, or the evaluator appropriately determines it based on some assumption. In either method, the measured data is once aggregated into a small number of statistical data, and the quality is evaluated based on the statistical data.

As a method for directly obtaining the line speed based on the original data of traffic, there is Japanese Patent Application No. 9-291051 (unpublished at the time of filing of this application) "Required Speed Estimator and Cell Loss Number Estimator". This is because an optical branching device is inserted in the line between ATM exchanges, the traffic (cell) on the line is monitored by the system attached to the end of the optical branching device, and the time stamp of each cell flowing through the line is used to check the line. It calculates a formula that gives the velocity, and the calculation result of the formula is obtained every time a certain number of cells flow.

That is, when traffic is measured in the ATM switch, the load on the switch increases. However, when traffic measurement is performed on the output line of the exchange, the cells are already multiplexed according to the transmission schedule, and correct measurement cannot be performed. Here, in this multiplexed line, when the arrival time at the observation device inserted for the first cell of a series of cell flows (here, one packet) is actually observed, statistical processing is performed from this actually observed value. It is a technique that can estimate the arrival time at the ATM switch outgoing line buffer.

[0007]

In order to carry out highly accurate quality evaluation for traffic design, the following points which are problems in the prior art are solved. (1) In the traffic measurement of the information communication system, there are many restrictions on items such as measurement time length, and much information about traffic is lost. (2) A large amount of information is lost by aggregating the traffic measurement data into a small number of statistical parameters in order to fit a specific probability / statistical model. The actual traffic is much more complicated than the stochastic / statistical model described by a small number of parameters, and the modeling error cannot be ignored. (3) When the line speed or the like is to be given by the mathematical formula, the approximate formula is often used, and the approximation accuracy is not necessarily high.

The present invention has been made against such a background, and an object thereof is to provide an evaluation apparatus capable of evaluating communication quality without losing information regarding traffic. It is an object of the present invention to provide an evaluation device capable of evaluating communication quality without being affected by modeling error. An object of the present invention is to provide an evaluation device that can evaluate communication quality without using an approximate expression.

[0009]

In order to solve the above problems, the present invention (1) obtains the traffic primitive data by inserting a branching device or using an access history. Do not use the traffic measurement items of the information communication system. (2) The obtained raw data is directly given to the simulation without being aggregated into statistical parameters. As a result, almost the actual traffic can be reproduced. In order to continuously perform such a simulation, the measurement point and the execution part of the simulation logic are connected online. (3) The quality is evaluated by simulation, and the mathematical formula is not used.

The above features are dealt with.

That is, the present invention is a communication quality evaluation apparatus which comprises means for branching and capturing a packet passing through a point of interest of an exchange, and observing the packet branched by this means to evaluate the quality corresponding to the traffic. is there.

Here, a feature of the present invention is a means for giving a time stamp of the arrival time at the point of interest for a specific processing unit forming a packet input from the fetching means, and this giving means. Means for estimating the arrival time at the point of interest for the processing unit included in the packet from the time stamp of the specific processing unit given by the means, and the arrival time at the point of interest estimated by the estimating means And means for simulating the arrival of the packet at the point of interest according to a preset simulation logic at a time corresponding to.

Assuming that the exchange is an ATM exchange and the point of interest is its outgoing line buffer, the processing unit is a cell and the specific processing unit is the first cell in the packet. it can. Alternatively, the processing unit may be a cell and the specific processing unit may be an individual cell forming the packet.

At this time, the means for simulating is
The packet may be configured such that the estimated arrival time of the i-th cell forming the packet is [time stamp of the first cell + (i−1) × cell length / incoming line speed], or the packet is The estimated arrival time of the i-th cell to be configured is [estimated arrival time of the first cell + (i−1) × cell length / incoming line speed], and the estimated arrival time of the first cell is the reference time and the first time. It is the sum of the estimated time required for the cell to arrive, which is output from the cell time stamp to the reference time.

Here, when a cell is output from the exchange to the line without a gap from the first cell of the previous packet to the first cell of the packet, the reference time is the first cell of the previous packet. The estimated arrival time, the reference time is the arrival time of the first cell of the previous packet, and when there is a gap from the first cell of the previous packet to the first cell of the packet, the reference time is set. Is equal to the reference time, and this can be configured to include means for determining the latest "pause" end time.

Alternatively, the means for simulating is
It has a table in which the estimated arrival cell rate for each time is recorded, and reads the estimated arrival cell rate a at the time indicated by the time stamp of the packet. If this value is 1 or more, the estimated arrival time t (1 ) According to [“previous packet first cell arrival time estimated value t ′ (1)” + (“corresponding packet time stamp t” − “previous packet time stamp t ′”) / arrival cell rate estimated value a], ” If the estimated arrival cell rate a is less than 1, the arrival time estimated value t (1) of the first cell of the packet is [[the first packet arrival time of the relevant packet] + “waiting time of the first cell of the packet” = “the relevant packet” First cell transmission time "=
It is also possible to adopt a configuration including means for calculating according to "the packet time stamp t".

If the exchange is a router and the point of interest is its outgoing line, the processing unit is a packet, and the transmission end of the i-th packet and the start transmission of the (i + 1) th packet continue. , The (i + 1)
The estimated arrival time of the packet is [estimated arrival time of the i-th packet + transmission start time of the (i + 1) th packet] / 2, and the transmission start time of the (i + 1) th packet is (i +
1) Time when [packet length of (i + 1) th packet / outgoing line speed] is subtracted from the arrival time of the packet, i.
If there is a gap between the end of packet transmission and the start of transmission of the (i + 1) th packet, the estimated arrival time of the (i + 1) th packet is the transmission start time of the (i + 1) th packet, and the estimated arrival time of the (i + 1) th packet is The transmission start time is the (i +
1) It can be configured to include a means for determining that it is a time obtained by subtracting [the packet length of the (i + 1) th packet / outgoing line speed] from the arrival time of the packet.

As a result, the communication quality can be evaluated without losing the traffic information. Also,
Communication quality can be evaluated without being affected by modeling error. Furthermore, it is possible to evaluate communication quality without using an approximate expression.

[0019]

BEST MODE FOR CARRYING OUT THE INVENTION An embodiment of the present invention will be described with reference to FIGS. FIG. 1 is a block diagram of the essential parts of a communication quality evaluation apparatus according to the first embodiment of the present invention. FIG. 2 is a block diagram of the essential parts of the ATM exchanges 101 and 117 which are the objects of traffic design. An embodiment of the present invention will be described with reference to a first embodiment of the present invention.

The present invention, as shown in FIG. 1, is an optical branching device 210 as a means for branching and capturing a packet passing through a line 109, which is a point of interest of ATM switches 101 and 117.
And a communication quality evaluation device 211 for observing packets branched by the optical branching device 210 and evaluating quality corresponding to traffic.

Here, a feature of the present invention is that the cell arrival detection unit 213 is a means for giving a time stamp of the arrival time at the line 109 for a specific cell in the packet input from the optical branching device 210. The CPU 215 estimates the arrival times at the line 109 for a large number of cells included in the packet from the time stamps of the specific cells provided by the cell arrival detection unit 213, and the CPU 215 further estimates the estimated times. Line 109
The arrival of the packet on the line 109 is simulated according to a preset simulation logic at a time corresponding to the arrival time.

In the first to third embodiments of the present invention, the specific cell is the head cell in the packet, and the cell arrival detection unit 213 determines the arrival time of the head cell to the line 109 as the first cell. Identified as the arrival time of.

[0023]

(First Embodiment) In the first embodiment of the present invention, FIG.
The speed of the line 109 between the ATM exchanges 101 and 117 and the cell output unit 10 which is an output buffer to the line 109
An invention for evaluating the quality (cell loss rate due to buffer overflow, etc.) when the number of buffers in 6 is changed and the traffic increases by a certain amount will be shown.

In FIG. 2 , ATM switches 101 and 117 have cell input sections 102, 103 and 112, 113, switch sections 104 and 114, and cell output sections 105, 106 and 115 and 116, respectively, and incoming lines 107 and 108, respectively.
And input cells from 111 and 109 are input to the cell input unit 10
Received from 2, 103, 112, 113, switch unit 1
04, 114, and cell output units 105, 106, 1
It outputs to the outgoing lines 109, 110, etc. via 15, 116. The cell output units 105, 106, 115 and 116 have
It has a physical layer function for output to a buffer and an outgoing line.

The evaluation system of the present invention is applied to the line 109. As shown in FIG. 1 , an optical branching device 210 is inserted on the line 109 to branch the traffic (cell) flowing through the line 109 as light. As a result, the cells on the line 109 flow unchanged, but the same cells are brought from the optical branching device to the evaluation device 211. The cell is subjected to processing such as termination of the physical layer in the cell input section 212 and reaches the cell arrival detection section 213. The cell arrival detection unit removes the payload of the arriving cell, adds a time stamp, and sends it to the final cell detection unit 214.

FIG. 3 is a block diagram of the essential parts of the final cell detector 214. As shown in FIG. 3, the final cell detection unit 214 includes a VPI / VCI / PTI detection unit 301, a register 302, and a read register 303. FIG. 4 is a diagram showing the configuration of the register 302. As shown in FIG. 4, the register 302 holds VPI / VCI / cell number / time stamp / active VC number / busy bit / reference time. However, the number of active VCs / busy
The bit / reference time is necessary in case 2 of FIG. 7 described later.
Option 1 is unnecessary. FIG. 9 shows the VPI / VCI / PTI detection unit 3
It is a flowchart which shows the processing procedure of 01. VP in the last cell detector that received the cell header with time stamp
The I / VCI / PTI detection unit 301 performs the processing shown in FIG. Only in the case of FIG. 7 plan 2, which will be described later, the processes of 918-920, the number of active VCs, the reference time, and the busy bit process are required. Therefore, the memory 216 always holds the number of active VCs, the reference time, and the current value of the busy bit.

The VPI / VCI / PTI detecting section 301 is
Whether or not cells remain in the buffer is detected by whether or not the time stamp difference between the previous cell and the current cell is the one cell transmission time difference (91).
8). At this time, in actuality, a time difference of fluctuation corresponding to the overhead of the physical layer may occur. If the time difference is not one cell, the value of the time stamp is set to the reference time of the memory 216, and the busy bit is set to "0" to indicate that the cell is not retained in the buffer (919). The "previous cell time stamp" is updated to detect the one cell time difference (920).

The VPI / VCI / PTI detector 301 is
Start detection of VPI / VCI / PTI. First, PT
Read the I area and indicate the last cell of the data packet 000,
When 010 is detected (901), VPI /
VCI is read (902), and if there is register data of the same VPI / VCI as the same VPI / VCI (903),
It is read (904). This corresponds to the final cell of a packet with two or more cells. The value obtained by adding “1” to the number of cells of the register data (905), the time stamp of the data of the register and the number of active VCs / busy bits / reference time are read out from the register 303.
Write to. The number of current active VCs is decremented by "1" (906). This read register is C
It is read by the PU 215. VPI / VCI
The register data of is deleted (907). If there is no register data (903), this corresponds to one cell in one packet, so the number of cells = 1 (908), and the time stamp given to the cell header and the current cur are set.
rent active VCs as the number of active VCs are written in the read registers respectively, and
The busy bit is set to "1" and the time stamp of the same cell is set as the reference time (909).

On the other hand, when PTI = 001, 011 (9
10) is a non-final cell in the data packet, so
Perform the corresponding processing. VPI / VC of cell header
I is read (911), and it is checked whether there is register data corresponding to the VPI / VCI (912). If there is, this is a non-leading (non-final) cell, so the value of the number of cells in the same data is added by "1" (917), and the processing ends. If there is no register data, this is the first cell (non-final cell), and a new entry is created in the register 302 (913 to 916). In particular,
VPI / VCI setting (915), cell number initialization (9
16), the value given to the same cell header is set as the time stamp value (913), and the current active VC is set as the number of active VCs, and c
"1" is added to the current active VC.
Further, the current busy bit and the reference time value are set in the busy bit area and the reference time area of the register 302. Then, the busy bit is set to "1" and the time stamp value of the cell is set as the reference time of the memory 216, thereby updating (914).

By the operation of the final cell detection unit 214, the read register 303 stores the time stamp and the number of cells (of the first cell), the number of active VCs, the number of active VCs, and the beginning of the same packet in which the last cell is detected as the last cell every time the last cell is detected. The busy bit value at the time of cell arrival and the reference time are written. The busy bit upon arrival of the head cell indicates whether or not the buffer has been emptied since the head cell of the immediately preceding packet. If it is empty, the reference time of the packet is the end time of the buffer empty immediately before the first cell of the packet, and if it is not empty, the first cell of the previous packet is sent out. It's time. This information is CP
It is read by U215 and used for quality evaluation as described below.

FIG. 5 is a flow chart showing the initialization procedure. First, as its preparation, initialization is performed as shown in FIG. First, the output line speed and the size of the buffer to be evaluated are input from the keyboard 220, and the CPU 215 inputs the input data via the I / O control unit 217.
It is read and stored in the memory 216 (501). Similarly, what kind of cell output section processing discipline is to be evaluated (usually some typical ones such as FIFO and HOL are prepared as options and selected from them) keyboard 220
(502). In the same way, the degree of traffic increase (100u%) compared to the current traffic is evaluated (503). The number of cells in the simulation buffer for quality evaluation is initialized (usually to 0) (504), and the simulation time is usually set to
[(Current time)-(Buffer size) of the evaluation system x
(Cell length) / (outgoing line speed)] (50
5). A cell counter 218 that collects statistical data for generating additional traffic is initialized (506), a cell counter upper limit value for resetting the cell counter is input from the keyboard 220, and the CPU 215 stores it in the memory 216 (507). ). The speeds (bandwidths) of the incoming lines 107 and 108 are similarly input and stored (508).

Further, FIG. 6 shows a process of acquiring statistical data for generating additional traffic. First, CPU2
15 is a timer and cell counter 218 for this processing
Is set to "0" (601, 602). The cell counter 218 counts the number of cells (603).
When the number of cells reaches N (604), the timer value at that time is read (605). This timer value corresponds to N cell arrival interval data. If M data are collected (60
6) Based on that, the CPU 215 calculates the average arrival interval A of cells and the standard deviation S (607).

A method for quality evaluation based on these is shown in FIGS. FIG. 7 relates to a method of generating a time-of-arrival sequence of cells. FIG. 8 shows a method of performing a simulation for performance evaluation based on the arrival time sequence.

Two schemes are shown in the arrival time sequence generation method of FIG. Option 1 is a relatively simple method, and act
The number of ive VCs, busy bit, and reference time are unnecessary. The plan 2 is a method with higher accuracy than the plan 1 especially when the load is heavy.

In plan 1, the CPU 215 reads out the number of cells n and the time stamp t from the final cell detection unit internal read register 303 (701). Based on this t and n, C
The PU 215 sets [t + (i) as an estimated value t (i) at the time when the packet arrives at the i-th cell for i = 1, ..., N.
-1) x (cell length) / incoming line speed] is calculated (70
2) The t (1), ..., T (n) are added to the elements of the arrival time sequence set and stored (703). The incoming line speed is stored in the memory 216 at 508.
The cell length is a fixed value and is stored in the memory 216 in advance.

In plan 2, the CPU 215 uses the read register 303 in the final cell detection unit to read the number of cells n, the time stamp t, the number of active VCs k, the busy bit b,
The reference time T is read (704). This t, n, k,
Based on b and T, the CPU 215 sets [T ′ + (outgoing line speed ratio) × (t−T) for i = 1, ..., N as the estimated arrival time t (i) of the i-th cell of the packet. ) / K ′
+ (I−1) × (cell length) / incoming line speed] is calculated (705). However, k ′ = max (1, k),
t '(1) is the estimated cell arrival time of the leading cell of the previous packet, and the reference time T'is the reference time T when b = 0 and b = 1.
Then t '(1). That is, when the buffer becomes empty, the next time when the buffer becomes empty is taken, and when not, the estimated arrival time of the leading cell of the previous packet is taken as the reference time.

The second term is the number of output cells during (t-T) / input cell rate, and represents the time required to input the number of cells output during (t-T). Therefore, the arrival time of the leading cell of the packet can be estimated by the first and second terms. The third term may be added to the arrival time of the subsequent cell in the packet. t '(1) is updated each time (706). 7
The t (1), ..., T (n) obtained in 05 are stored in addition to the arrival time sequence set.

Proposals 1 and 2 are both cell arrival time sequences for the current traffic. In addition to this, the generation of the cell arrival time sequence 707 to 709 when the traffic increases
Done in. This process increases traffic (u> 0)
Only when is, it is started (707). In the process of FIG. 6, the average A of the current cell arrival intervals and the standard deviation S are obtained at regular intervals. The average is 10 while maintaining the coefficient of variation S / A.
A random number x having an average A / u of arrival intervals and a standard deviation S / u is generated so as to add 0 u%, and this random number x is used as an incremented cell arrival interval (708). The random number is S / A>
If 1, then H2 distribution, if S / A = 1, then Poisson distribution,
If S / A <1, it is possible to easily generate random numbers having desired statistical characteristics by using the Erlang distribution. Generated random number x
Is set as the arrival interval, and x is added to the arrival time y of the previous cell to obtain the arrival time of the next cell (709). This is stored in addition to the arrival time sequence set (703).

According to the procedure of FIG. 7, the CPU 215 has a group of arrival time sequence data which is updated every moment.
A method of performing a quality evaluation simulation based on the data will be described with reference to FIG. However, here, 502 in FIG.
In the above description, the FIFO is selected as the cell output processing discipline. The method when another processing discipline is selected can be easily inferred, so I will not describe it, but since it is necessary to facilitate the simulation method in advance to support the processing discipline, the processing discipline must be prepared in advance. You will be selected from the ones that have been given.

The CPU 215 selects the closest arrival time from the arrival time sequence set as the next arrival time (80).
3). When the arrival time is reached (804), the number of arriving cells is incremented by 1 (805). If there is no cell number in the buffer (814), the cell transmission time of the arrival cell is set as the next cell transmission time, that is, [arrival time + (cell length) /
(Outgoing line speed)] is set (815). When the arriving cell overflows the buffer (806), the cell loss number is set to 1
If the buffer is not overflowed (808), the integrated value of the waiting time is added and the number of cells in the buffer is added by "1" (807).

When the cell transmission time is reached (809), the number of cells in the buffer is decremented by "1" (810). As a result, when there are no more cells in the buffer (811), the cell transmission time is set to a sufficiently large value (813). This value is updated when the cell arrives. When there are still cells in the buffer (811), as the cell transmission time, the transmission time of the next cell [current cell transmission time + (cell length) / (outgoing line speed)] is set to the cell transmission time (812). ). The cell loss number, the waiting time integrated value, and the number of arriving cells obtained as a result of the simulation are I / O control units 217 at regular intervals.
After that, it is output to the monitor 219 (802). If necessary, the cell loss rate (= cell loss number / arrival cell number) and average waiting time (= waiting time / (arrival cell number-cell loss number)) are converted and output.

In both plan 1 and plan 2, the last cell is detected by the last cell detection unit 214, and the first cell is detected as a result, so that the output line buffer of each cell is based on the time stamp of each packet. Is estimating the arrival time. On the other hand, the cell arrival detection unit 213 directly
It is conceivable that the arrival time of each cell at the outgoing line buffer is estimated by the time stamp of each cell given by.

Normally, when the time stamp of each cell is used, the final cell detection unit 214 is unnecessary. The easiest method for estimating the arrival time of each cell at the outgoing line buffer by the time stamp of each cell is to use the time stamp of each cell as the estimated arrival time at the outgoing line buffer.

That is, the CPU 215 reads the time stamp of the cell arrival detection unit 213 for each cell and uses this as the arrival time sequence. The generation of the additional traffic arrival time sequence is the same as the procedure described above.
Based on this, the communication quality evaluation simulation can be performed according to the procedure described above.

(Second Embodiment) The second embodiment of the present invention is similar to the first embodiment of the present invention, and the ATM switch 10 shown in FIG.
The speed of the line 109 between 1 and 117 and the number of buffers in the cell output unit 106, which is an output buffer to the line 109, are changed to evaluate the quality (cell loss rate due to buffer overflow) when the traffic increases by a certain amount. Indicates the invention.

However, in the second embodiment of the present invention, the time stamp and bu for each cell required in the first embodiment of the present invention are added.
The difference is that processing such as updating the sy bit may be performed for each packet.

A second embodiment of the present invention will be described with reference to FIGS. FIG. 10 is a diagram showing a table of arrival cell rate estimated values for respective times. FIG. 11 is a flowchart showing a procedure for generating a cell arrival time sequence according to the second embodiment of the present invention. FIG. 12 shows the VP of the second embodiment of the present invention.
It is a flow chart which shows a processing procedure of an I / VCI / PTI detection part.

The VPI / VCI / PTI detection unit 301 in the final cell detection unit which has received the time stamped cell header
Performs the processing shown in FIG. Register 302 is a VPI
In addition to holding / VCI / cell number, as shown in FIG. 10, it has a table of arrival cell rate estimated values for each time.

The VPI / VCI / PTI detecting section 301 is
Start detection of VPI / VCI / PTI. First, PT
Read the I area and indicate the last cell of the data packet 000,
When 010 is detected (1901), the VPI of the same cell
/ VCI is read (1902) and if there is register data of the same VPI / VCI as the same VPI / VCI (190
3) Read it (1904). This corresponds to the final cell of a packet with two or more cells. "1" is added to the number of cells of the register data (1905), and the time stamp which was the data of the register is written in the read register 303 (1906). This read register is read by the CPU 215. VPI / V
The CI register data is deleted (1907). If there is no register data (1903), this is 1
Since one packet corresponds to one cell, the number of cells = 1 (190
8), the time stamp added to the cell header is written in the read register (1909).

On the other hand, when PTI = 001, 011 (1
Since 910) is the non-final cell in the data packet, the corresponding process is performed. VPI of cell header /
The VCI is read (1911), and it is checked whether there is register data corresponding to the VPI / VCI (191).
2). If there is, this is a non-leading (non-ending) cell, so the value of the number of cells in the same data is added by "1" (191
7) End the process. If there is no register data, this is the first cell (non-final cell), so a new entry is created in the register 302 (1913 to 1913).
15). Specifically, VPI / VCI setting (191
4), the number of cells is initialized (1915), and the value given to the same cell header is set as the time stamp value (1
913). Here, in the first embodiment of the present invention, activ
e The number of VCs, the busy bit area, and the reference time area are required, but they are not required in the second embodiment of the present invention.

By the operation of the final cell detection unit 214, the time stamp of the packet having the same cell as the final cell (the time stamp of the first cell of the packet) and the number of cells are written in the read register 303 every time the final cell is detected. Be done. In the above method, the number of cells forming the packet is specified from the ATM cell header, but the number of cells forming the packet may be determined from the packet length described at a predetermined position in the payload of the first cell.

Next, quality evaluation is performed from the time stamp of each packet and cell number information. First, as its preparation, initialization is performed as shown in FIG. First the outgoing line speed,
What kind of buffer size you want to evaluate,
Input from the keyboard 220, read by the CPU 215 via the I / O control unit 217, and stored in the memory 216 (501). Similarly, what kind of cell output section processing discipline should be evaluated (usually, some typical ones such as FIFO and HOL are prepared as options and selected from among them).
The selection is made using the keyboard 220 (502). How much traffic increases from the current traffic (10
0u%) is evaluated or similarly input (503). The number of cells in the buffer of the simulation for quality evaluation is initialized (normally to 0) (504), and the time of the simulation is normally set to [current time of evaluation system- (buffer size) x (cell length) / ( (Outgoing line speed)] (505). Initialize (506) a cell counter 218 that collects statistical data for generating additional traffic,
The cell counter upper limit value for resetting the cell counter is input from the keyboard 220, and the CPU 215 stores it in the memory 216 (507). Incoming lines 107, 10
Similarly, the speed (bandwidth) of 8 is input and stored (508).

Further, FIG. 6 shows a process of acquiring statistical data for generating additional traffic. In the following, the average and standard deviation of cell arrivals are obtained from the cell counter, but the same method can be used depending on the number of cells forming the packet and the time stamp of the packet. First, CPU2
15 is a timer and cell counter 218 for this processing
Is set to "0" (601, 602). The cell counter 218 counts the number of cells (603).
When the number of cells reaches N (604), the timer value at that time is read (605). This timer value corresponds to N cell arrival interval data. If M data are collected (60
6) Based on that, the CPU 215 calculates the average arrival interval A of cells and the standard deviation S (607).

A method for quality evaluation based on these is shown in FIG.
1 and FIG. In the arrival time sequence generation method of FIG. 11,
The CPU 215 reads the number of cells n and the time stamp t from the read register 303 in the final cell detection unit 214 (1701). As shown in FIG. 10, the CPU 215 holds a table of arrival cell rate estimated values for respective times. The arrival cell rate estimation value a at the time indicated by the time stamp of the packet is read, and if this value is 1 or more (1702), the arrival time estimation value t (1) of the first cell of the packet is set to "[previous packet first cell Arrival time estimated value t ′ (1) ”+“ arrival time difference between the first cell of the previous packet and the first cell of the packet ”].

"The first cell of the previous packet and the first cell of the packet
The “cell arrival time difference” is based on the assumption that the output link of the previous packet first cell does not become empty during the transmission of the packet first cell, [(“the packet time stamp t”
-"Previous packet time stamp t '" / arrival cell rate estimated value a]. That is, [t (1)
= "Previous packet first cell arrival time estimated value t '(1)" +
(“The packet time stamp t” − “previous packet time stamp t ′”) / arrival cell rate estimated value a] (1
703). If the estimated arrival cell rate a at the time indicated by the time stamp is less than 1 (1702), the first packet
The estimated arrival time t (1) of the cell is [[
Cell arrival time ”+“ waiting time for the first cell of the packet ”
= "The packet first cell transmission time" = "the packet time stamp t"]. The estimated value w of the “waiting time of the first cell of the packet” is, for example,
Average waiting time of M / M / 1 queue with arrival rate a and 95%
Use points, etc. In FIG. 11, cell length / (outgoing line speed × (1-a)), which is the average waiting time of the M / M / 1 queue.
Is used (1704). Therefore, t
(1) = “corresponding packet time stamp t” −w (1704). The previous packet time stamp and the previous packet first cell arrival time estimated value are updated (1705).

The time required for the n cells forming the packet to arrive from the incoming line is n × (cell length) / incoming line speed. Based on this, the time t is compared with the arrival cell rate estimated value table shown in FIG.
The arrival cell rate estimation value between (1) to t (1) + n × (cell length) / incoming line rate is increased by the incoming line rate / cell length (1706).

After a predetermined time delay, cells are randomly generated according to the estimated arrival cell rate at each time. As a result, a cell arrival time sequence is generated.

In addition to this, the generation of a cell arrival time sequence in the case of a traffic increase is performed in steps 1707 to 170.
It will be held between nine. This process increases traffic (u
Only when> 0), it is activated (1707). In the process of FIG. 6, the average A of the current cell arrival intervals and the standard deviation S are obtained at regular intervals. Average A / u of arrival intervals so that the average is added by 100 u% while keeping the coefficient of variation S / A,
A random number x having a standard deviation S / u is generated, and this random number x is used as an increment cell arrival interval (1708). Random numbers are
If S / A> 1, H2 distribution is used, if S / A = 1, Poisson distribution is used, and if S / A <1, Erlang distribution is used. Random numbers having desired statistical characteristics can be easily generated. The generated random number x is set as the arrival interval, and x is added to the arrival time y of the previous cell to obtain the arrival time of the next cell (1709). This is stored in addition to the arrival time sequence set (1703).

According to the procedure of FIG. 11, the CPU 215 has a group of arrival time sequence data which is updated every moment. Similar to the first embodiment of the present invention, a quality evaluation simulation is performed based on the data shown in FIG.

(Third Embodiment) FIG. 13 shows a block configuration of essential parts of a router and a traffic design quality evaluation apparatus according to a third embodiment of the present invention. In the third embodiment of the present invention, as shown in FIG. 13, a line 1008 between routers 1001 and 1010,
The quality of 1009 traffic is evaluated. Network interface units 1002, 1003, 1006, 100 that perform lower layer processing associated with the termination of a line, and buffer and transfer processing of packets
The router 1001 having No. 7 sends out the packet input from each network interface unit to the line via the corresponding network interface unit according to the routing table managed on the memory 1005.
Also, according to the control information and routing protocol,
The CPU 1004 changes the routing table in the memory 1005 as necessary. Line speed 100
If the buffer amount of 8, 1009 or the network interface unit is insufficient, a delay due to retention of packets in the output network interface unit or packet discard due to buffer overflow occurs.

The branch device 1017 is connected to the desired line 1008,
1009, and duplicately monitors the packet electronically or optically. The monitored packet is processed by the network interface unit 1019 in the lower layer and sent to the packet arrival detection unit 1020. The packet arrival detection unit removes the payload, removes the packet length information written at a predetermined position defined by the packet format, removes the packet header, and adds a time stamp. The CPU 1021 reads the packet length and time stamp thus obtained. CPU 1021
Uses this information to evaluate the quality of the traffic on the lines 1008, 1009. First, as its preparation, initialization as shown in FIG. 5 is performed. It can be easily inferred by replacing the cell in the figure with a packet, and the description thereof will be omitted.

Further, FIG. 14 shows a process of obtaining statistical data for generating additional traffic. First, the CPU 1021 resets the timer for this processing, the packet counter 1026, and the packet length counter 1027 (1101, 1102). Packet arrival detection unit 1020
When the CPU 1021 finds new packet data in, the CPU 1021 adds the value of the packet counter 1026 to "1" and the value of the packet length counter 1027 by the amount of packet length data found by the packet arrival detection unit (1103). . When the number of packets reaches N ((11
04), the CPU 1021 reads the timer value and the packet length counter value and stores them in the memory 1022 (1105). When M timer values (and packet length counter values) are stored (1106), the CPU 1021 causes the memory 10 to store the average arrival interval A of packets, the standard deviation S of arrival intervals, the average value LA of packet lengths, and the standard deviation LS.
It is calculated from the timer value and the packet length counter value stored in 22 (1107).

The traffic on the lines 1008 and 1009 is reproduced, and the quality is evaluated by simulation. The time stamp obtained from the packet arrival detection unit 1020,
FIG. 15 shows a method in which the CPU 1021 estimates the arrival time of the packet from the packet length information. There is also a simple method (the accuracy is deteriorated when the load is high) in which [timestamp- (packet length) / (outgoing line speed)] is used as it is as an estimated value of the arrival time.

The CPU 1021 uses the time stamp value of the next packet when the packet is accumulated in the buffer of the output network interface unit- (next packet length) /
(Outgoing line speed). Next time stamp scheduled time,
The [timestamp- (packet length) / (outgoing line speed)] calculated from the latest time stamp and packet length obtained from the packet arrival detection unit is compared, and if they match (1201), this is stored in the buffer. Since the packets are accumulated, the estimated arrival time of the packet is the "previous arrival time" which is the estimated arrival time of the immediately preceding packet and [timestamp- (the packet length) / (outgoing line speed)].
An intermediate time of is used (1202). At this time, in practice, the difference for the predetermined fluctuation does not matter. Otherwise, the packet is transmitted without staying in the buffer, so [timestamp- (packet length) / (outgoing line speed)] is used as the estimated arrival time (1203). The "previous time" and "scheduled next time stamp" are updated (1204). The additional traffic is generated by a random number based on the statistical data obtained in FIG. 14 as in the first embodiment of the present invention (1205). The packet length is also the same (1206), and the arrival time y of the additional traffic
Is changed (1207).

Thus, based on the packet obtained each time the packet arrives, the estimated arrival time and the packet length information, the CPU
FIG. 16 shows a method in which the 1021 performs quality evaluation by simulation. However, the processing discipline will be described assuming that the FIFO is selected. Even if another discipline is selected, it can be easily analogized.

When the packet arrival time obtained in FIG. 7 is reached (1302), the counter of the number of arrived packets is incremented by "1" (1303). If there is no packet in the simulation buffer at that time (1304), the next packet transmission completion time is the packet transmission completion time of the packet that has arrived now. [Arrival time + (packet length)
/ (Outgoing line speed)] is set (1305). Now, determine whether the packet that has arrived will cause a buffer overflow,
If it occurs (1306), the packet loss counter is incremented by "1" (1307). If it does not occur, add [(total packet length in buffer) / (outgoing line speed)], which is the waiting time of the packet, to the integrated value of the waiting time, and add the packet length of the packet to the total packet length in the buffer. Update. Further, the packet length of the packet is added to the packet length list (1308). When the packet transmission completion time is reached (1309), the first element, which is the packet length of the packet that has been transmitted, is removed from the packet length list, and the total packet length in the buffer is reduced by the packet (1310). As a result, when the packet length list becomes empty (1311), a sufficiently large value is set for the next packet transmission completion time (1312).
It is updated with the next packet arrival. If the packet length list is not empty, the next packet transmission completion time is set to the transmission completion time when the packet is transmitted subsequently to the previous packet transmission completion time (1313).
When a certain output time is reached (1301), the integrated values of the number of arrived packets, the waiting time, and the packet loss are calculated by the I / O control unit 1.
Output to the monitor 1024 via 023 (131
4). If necessary, [packet loss rate = number of packet losses / number of arriving packets] and [average waiting time = waiting time / (number of arriving packets−number of packet losses)] are calculated and output.
After the output, the number of arriving packets, the number of packet losses, and the waiting time are reset (1315).

(Summary of Embodiments) In the first to third embodiments of the present invention, the branching device is inserted on the line and the time stamp data of the cell or packet on the line is directly obtained. Therefore, there is no loss of information due to the measurement items in the ATM switch or router. Also, the arrival time of the cell or packet is estimated from the time stamp of each cell or packet, and is directly given to the simulation as data.
Therefore, there is no information loss or modeling error due to aggregation into a small number of statistical data. In addition, the evaluation accuracy is high because the simulation is performed regardless of the mathematical formula.

[0068]

As described above, according to the present invention,
It is possible to evaluate communication quality without losing information about traffic, without being affected by modeling error, and without using an approximate expression.

[Brief description of drawings]

FIG. 1 is a block diagram of a main part of a traffic design quality evaluation device according to an embodiment of the present invention.

FIG. 2 is a block diagram of a main part of an ATM exchange that is a traffic design target.

FIG. 3 is a block diagram of a main part of a final cell detection unit.

FIG. 4 illustrates a structure of a register.

FIG. 5 is a flowchart showing an initialization procedure.

FIG. 6 is a flowchart showing a process of acquiring statistical data for generating additional traffic.

FIG. 7 is a flowchart showing a procedure for generating a cell arrival time sequence.

FIG. 8 is a flowchart showing a procedure for performing a simulation for performance evaluation based on the arrival time sequence.

FIG. 9 is a flowchart showing a processing procedure of a VPI / VCI / PTI detection unit.

FIG. 10 is a diagram showing a table of arrival cell rate estimated values for respective times.

FIG. 11 is a flowchart showing a procedure for generating a cell arrival time sequence according to the second embodiment of the present invention.

FIG. 12: VPI / VCI / PTI of the second embodiment of the present invention
The flowchart which shows the process sequence of a detection part.

FIG. 13 is a diagram showing a block configuration of main parts of a router and a traffic design quality evaluation device according to a third embodiment of the present invention.

FIG. 14 is a flowchart showing a procedure of an acquisition process of statistical data for generating additional traffic.

FIG. 15 is a flowchart showing a procedure for the CPU to estimate the arrival time of a packet.

FIG. 16 is a flowchart showing a procedure in which the CPU performs quality evaluation by simulation.

[Explanation of symbols]

107-111, 1008, 1009 lines 101, 117 ATM switch 102, 103, 112, 113 Cell input section 104, 114 Switch section 105, 106, 115, 116 Cell output section 210 Optical branching device 211, 1018 Evaluation device 212 cell input section 213 Cell arrival detector 214 Final Cell Detection Unit 215, 1021 CPU 216, 1022 memory 217, 1023 I / O control unit 218 cell counter 219, 1024 monitor 220 keyboard 301 VPI / VCI / PTI detector 302 register 303 Read register 1001, 1010 router 1019 Network interface section 1017 Branch device 1020 packet arrival detection unit 1026 packet counter 1027 packet length counter

─────────────────────────────────────────────────── ─── Continuation of the front page (58) Fields surveyed (Int.Cl. ⁷ , DB name) H04L 12/56 H04L 12/24 H04L 12/26

Claims (12)

(57) [Claims] 1. A packet passing through a point of interest of a switch is divided.
It is equipped with a means to diverge and take in, and a packet branched by this means.
To evaluate the quality corresponding to traffic by observing trafficIn order
OutfitSettingAnd, Specifics that make up the packet input from the capturing means
Time stamp of arrival time at the point of interest for processing unit
Means for applying a pump, Of the specific processing unit given by the giving means.
Processing unit included in the packet from the time stamp
A means for estimating the arrival time at the point of interest for Arrival at the point of interest estimated by this estimating means
At timeRelated information andPre-configured simulator
SolutionTraffic simulation based on conditions
Means to carry out Equipped with The means for carrying out the simulation evaluates the communication quality.
Includes means for performing calculations to evaluate It is characterized by
Beliefsimulationapparatus. 2. The means for carrying out the simulation comprises:
Quality evaluation is performed as a condition for conducting the simulation.
Desired output line speed, buffer size, cell output
Estimate arrival time by physical discipline and the estimation means
Traffic to be added to the traffic when
The communication simulation apparatus according to claim 1 , further comprising means for determining the communication. 3. The means for performing the calculation is a packet loss number,
Packet waiting time accumulated value, number of arriving packets, or
Outputs packet loss rate and average packet waiting time
Communication simulation sheet according to claim 1 or 2, wherein including means that
® down apparatus. 4. The means for performing the calculation is cell loss number, cell
Waiting time integrated value, number of arriving cells, or cell loss rate,
Communication simulation apparatus according to claim 1 or 2, wherein including means for calculating outputs an average waiting time. 5. The exchange is an ATM exchange and the destination
Eye point communication according to claim 1, wherein its output line buffer sheet
Simulation device. 6. The processing unit is a cell, and the specific processing
The communication simulation device according to claim 5 , wherein the physical unit is the first cell that constitutes the packet . 7. The processing unit is a cell, and the specific processing is performed.
Claim management unit is each cell constituting the packet 5
The communication simulation device described. 8.The means for estimating is The estimated arrival time of the i-th cell forming the packet is [Timestamp of the first cell + (i−1) × cell length
/ Incoming line speed] Including means to be Claim5Listed communicationSimulation
Optionapparatus. 9.The means for estimating is The estimated arrival time of the i-th cell forming the packet is [Estimated arrival time of the first cell + (i-1) x cell length / entry time
Linear velocity) age, The estimated arrival time of this first cell is based on the reference time and the first cell.
The cell output from the time stamp of the
It is the sum of the estimated time it took to arrive, where: From the first cell of the previous packet to the first cell of the packet
The cells are output from the exchange to the line without any gap between
When The reference time is the estimated arrival time of the first cell of the previous packet.
Ticks, The reference time is the time stamp of the first cell of the previous packet.
The And where From the first cell of the previous packet to the first cell of the packet
And when there is a gap, The reference time is equal to the reference time, which is the latest
Including a means to be the end time of "between" Claim5Description
Communicationsimulationapparatus. 10.The estimation means is for each time
Equipped with a table recording the estimated arrival rates Arrival cell rate at the time indicated by the packet timestamp
Read the estimated value a and if this value is 1 or more,
Estimated arrival time t (1) of 1 cell “Previous packet first cell arrival time estimated value t ′ (1)” + ("The packet time stamp t"-"previous packet
Im stamp t '") / estimated arrival rate a And the arrival cell rate estimation value a is 1
If it is less than the estimated arrival time t of the first cell of the packet
(1) "The arrival time of the 1st cell of the packet" + "The arrival of the packet
Waiting time for one cell " = "Transmission time of the first cell of the packet" = "The relevant packet time stamp t" 7. Communication according to claim 6, comprising means for calculating according to
Simulation device. 11. The point of interest, wherein the exchange is a router.
The communication simulation according to claim 1, wherein is the outgoing line
Device. 12. The method according to claim 12,The processing unit is a packet, The above
The means to estimate is End of transmission of i-th packet and transmission of (i + 1) th packet
When the start is continuous, The estimated arrival time of the (i + 1) th packet is [Estimated arrival time of i-th packet + (i + 1) th packet
Transmission start time] / 2, The transmission start time of the (i + 1) th packet is From the arrival time of the (i + 1) th packet Subtract the packet length / outgoing line speed of the (i + 1) th packet
It is the time when End of transmission of i-th packet and transmission of (i + 1) th packet
When there was a gap between the start, The estimated arrival time of the (i + 1) th packet is the (i + 1) th packet.
Is the time to start sending the The transmission start time of the (i + 1) th packet is From the arrival time of the (i + 1) th packet Packet length / outgoing line speed of the (i + 1) th packet 12. The method according to claim 11, further comprising means for determining that the time is obtained by subtracting
Communication simulation apparatus.