JP3605771B2 - Polling time interval sequential correction system and method, recording medium on which processing program is recorded, and measuring device equipped with the system - Google Patents

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a technology for evaluating network performance, and in particular, a measurement accuracy for measuring the throughput of data communication passing through a specific communication device and the CPU usage rate of the communication device by polling from a remote place. The present invention relates to a system and method for successively correcting polling time intervals, which is suitable for improving the performance, a recording medium on which a processing program is recorded, and a measuring apparatus having the system.
[0002]
[Prior art]
Conventionally, as a technique for measuring the throughput of data communication through a specific link, a flow counter of a communication device located at any end of the link is acquired a plurality of times at a fixed time interval, and a difference between the flow counters is obtained for the time. In some cases, the throughput is obtained by dividing by an interval.
[0003]
In such a technique, not only the throughput but also a counter value that increases with time, and if the rate of increase per unit time is an amount to be measured, measurement is performed in exactly the same way. Is performed.
[0004]
Here, the flow rate counter is a counter that sequentially increases the value according to the amount of data that has passed through the link. As a standard, a MIB (Management Information Base) information of an SNMP (Simple Network Management Protocol) architecture is used. Of ifIn0ctets.
[0005]
In this SNMP, for example, by performing SNMP polling once every five minutes, the band of data flowing through the link can be measured from a remote place.
[0006]
However, in such a technique of acquiring the value of the flow counter by polling from a remote place, the time interval for acquiring the flow counter is a fixed value, and there is a problem in measurement accuracy for the following reasons.
[0007]
{Circle around (1)} The time interval for obtaining the flow counter value is the minimum unit of measurement time, and it is preferable that the time interval be short from the viewpoint of measurement time accuracy. However, if the time interval is set too small, the time interval at which the polling request packet actually reaches the communication device deviates from the actual polling time interval due to response delay fluctuation when acquiring the counter value by polling from a remote location. As a result, the measurement accuracy deteriorates. Therefore, it is necessary to perform polling at a time interval as short as possible and to determine a polling time interval that satisfies the required measurement accuracy.
[0008]
{Circle around (2)} In a network such as the Internet, such response delay fluctuation cannot be known in advance, and in the related art, since the polling interval is a fixed value, the accuracy of throughput measurement cannot be suppressed to a certain level. For example, in a low-speed network, the fluctuation of the response delay may reach as much as 5 seconds. In this case, if the polling interval is set to 30 seconds, the actual time interval for acquiring the counter value has an error of about 15%. Occurs. Therefore, it is desirable to increase the polling time interval to such an extent that the response delay fluctuation caused by the network can be ignored.
[0009]
[Problems to be solved by the invention]
The problem to be solved is that the conventional technique cannot make the polling time interval for obtaining the flow counter value too small.
[0010]
An object of the present invention is to solve these problems of the prior art and to improve the measurement accuracy of the rate of increase per unit time, such as the throughput of a predetermined link by polling at a remote location, and to improve the accuracy of the polling time intervals. It is an object of the present invention to provide a correction system and method, a recording medium on which a processing program is recorded, and a measuring device having the system.
[0011]
[Means for Solving the Problems]
In order to achieve the above object, the polling time interval successive correction system and method according to the present invention reduces the response delay time of the communication apparatus to polling up to the present by Ti (i = 1) n), the magnitude of the response delay fluctuation of the polling up to the present time is statistically evaluated from these amounts, and the time interval of the subsequent polling is determined based on the magnitude of the response delay fluctuation. decide. In this way, the polling time interval is determined with the smallest possible time interval and with the required measurement accuracy. In this way, the measuring apparatus of the present invention performs the measurement based on the polling at the time intervals determined according to the polling response delay fluctuation that changes depending on the network condition.
[0012]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
FIG. 1 is a block diagram showing a configuration example of a polling time interval sequential correction system according to the present invention and a throughput measuring device having the same, and FIG. 2 is a block diagram of a network to be measured by the throughput measuring device in FIG. FIG. 5 is a block diagram showing an example of a hardware configuration of a throughput measuring device provided with the measuring device polling time interval sequential correction system in FIG.
[0013]
5, reference numeral 51 denotes a display device such as a CRT (Cathode Ray Tube) or an LCD (Liquid Crystal Display); 52, an input device including a keyboard and a mouse; 53, an external storage device including a HDD (Hard Disk Drive); , 54 are an information processing apparatus having a CPU (Central Processing Unit) 54a and a main memory 54b for performing computer processing, and 55 is a CD-ROM (Compact Disc-Read Only Memory) recording programs and data according to the present invention. Alternatively, an optical disk such as a DVD (Digital Video Disc / Digital Versatile Disc) or the like, and 56 stores programs and data recorded on the optical disk 55 Drive for out look, 57 is a communication device comprising a LAN (Local Area Network) card or a modem.
[0014]
After the program and data stored in the optical disk 55 are installed in the external storage device 53 by the information processing device 54 via the driving device 56, the information is read from the external storage device 53 into the main memory 54b and processed by the CPU 54a. The throughput measuring device 1 including the polling time interval sequential correction system shown in FIG.
[0015]
In FIG. 1, a throughput measuring device 1 includes a polling packet transmitting unit 1a, a response packet (counter value) receiving unit 1b, a polling response delay time measuring unit 1c, a response delay data storage memory 1d, a polling response delay fluctuation calculating unit 1e, a polling It comprises an interval determining unit 1f, a polling schedule unit 1g, and a bandwidth calculating unit 1h, among which a polling response delay time measuring unit 1c, a response delay data storage memory 1d, a polling response delay fluctuation calculating unit 1e, a polling interval determining unit 1f, The polling schedule section 1g constitutes the polling time interval sequential correction system 2.
[0016]
As described above, by providing the polling time interval sequential correction system 2, the throughput measuring apparatus 1 of the present embodiment suppresses the measurement error to a certain level or less and reduces the throughput of data communication passing through a specific link. A measurement can be made.
[0017]
Here, the measurement target is a specific link 23 of the communication network 20 as the IP network shown in FIG. 2, and among routers (described as “nodes” 21 and 22 in the figure) 21 and 22 located at both ends of the link 23. By acquiring the MIB counter (ifInOctets) 24 of the router 21 by SNMP, the throughput measuring device 1 of the present example measures the throughput of the link 23.
[0018]
In the throughput measuring apparatus 1, each time polling is performed, in the polling time interval sequential correction system 2, the response delay time is measured by the polling response delay time measuring section 1c and written into the response delay data storage memory 1d, and the write result record is recorded. The polling response delay fluctuation calculator 1e sequentially calculates the polling response delay fluctuation, and the polling interval determiner 1f calculates a time interval until the next polling according to the magnitude of the polling response delay fluctuation.
[0019]
Hereinafter, details of such an operation will be described.
The throughput measuring device 1 measures the instantaneous value of the throughput at time intervals as short as possible. For example, polling is basically performed every 30 seconds. The data transfer amount during that period is the difference between the counter values obtained by two adjacent polls, and the measured throughput value is obtained by dividing the difference by 30 seconds.
[0020]
At this time, depending on the network (communication network 20), the time interval at which these two polling request packets reach the router (router 21) may be significantly different from 30 seconds. In such a situation, the throughput measuring device 1 of the present embodiment automatically sets the shortest polling time interval while maintaining the measurement accuracy.
[0021]
Such a polling response delay time is the time from when the throughput measuring device 1 sends a request packet to the router 21 by the polling packet sending unit 1a to when the response packet returns to the response packet (counter value) receiving unit 1b. Is defined as The polling response delay time Ti (i = 1... N) obtained here is recorded together with a time stamp obtained by the clock function of the throughput measuring device 1 when issuing a request packet.
[0022]
Assuming that this time is ti, the information accumulated after the i-th polling is {ti, Ti}. This information is obtained by the polling response delay time measuring unit 1c and stored in the response delay data storage memory 1d. The time ti is used to select Ti for a certain period in the past.
[0023]
The polling response delay fluctuation calculator 1e calculates the magnitude of the response delay fluctuation using the following statistical evaluation procedure. First, the response delay data of a certain period in the past counted from the present time is taken out from the response delay data storage memory 1d. It is assumed that there are m pieces of corresponding data, and they are set as Ti (i = 1... M). The next quantity, which is the difference between these two adjacent data of Ti, is calculated.
[0024]
That is, Xi = Ti−T _i−1, (i = 2... _M ). These values of Xi are often distributed in an IP network generally in a form as shown in FIG.
[0025]
FIG. 4 is a histogram showing an example of a distribution of a difference between response delay times in an IP network.
[0026]
If the standard deviation σ of Xi is taken as the response delay fluctuation, the width becomes approximately the distribution width shown in FIG.
The polling response delay fluctuation calculator 1e in FIG. 1 sets the width σ of this distribution as the value of the response delay fluctuation.
[0027]
Based on the value of the response delay fluctuation, the polling interval determination unit 1f calculates the time until the next polling as follows. That is, when the polling response delay fluctuation is σ, the time interval until the next polling is β · σ. Here, the value of β is a parameter mainly determined from the required measurement accuracy.
[0028]
If β> 1, the polling response delay error is reduced, and the throughput measurement accuracy is increased. However, if β is too large, the number of samples of the throughput measurement value obtained within a certain period decreases, so the polling interval determination unit 1f determines the parameter β as described below. Then, the polling schedule unit 1g sets a schedule so as to perform polling after this time interval.
[0029]
A specific procedure for determining a time interval until the next polling from the response delay fluctuation σ evaluated based on Xi (difference between two adjacent data) will be described below.
[0030]
The inputs are the difference Xi (i = 1... M) between two adjacent data, the required accuracy ε (relative error) of the throughput measurement, and the reliability p (a constant parameter input in advance). The output is the minimum polling time interval that satisfies the required measurement accuracy.
[0031]
Step (1) : Xi is calculated from the accumulated Ti data, and its standard deviation σ is calculated.
Step (2) : From the accumulated Xi data, calculate β ′ such that Prob ｛−β ′ · σ ≦ Xi ≦ β ′ · σ ｝ = p.
Step (3) : S = βm · σ is calculated based on βm represented by βm = (2 / ε + 1) β ′. This is the minimum polling interval required.
[0032]
The procedure (2) is a procedure for obtaining the range of the parameter β included in this distribution with the reliability p based on the accumulated distribution of Ti. Given a specific reliability p, the polling time interval will fluctuate between (β−β ′) · σ and (β + β ′) · σ .
[0033]
Here, since the measured throughput is proportional to the reciprocal of β · σ, the throughput fluctuates between V1 = 1 / (β−β ′) and V2 = 1 / (β + β ′), respectively. Therefore, if the value of the parameter β is set to be larger than β determined from the condition of (V2−V1) / V1 = ε, it is possible to satisfy the demand for the error of the throughput.
[0034]
Solving this for β yields βm = (2 / ε + 1) β ′.
If the time interval until the next polling is βm · σ based on ε and the reliability p as the required conditions, the error condition can be satisfied and the throughput measurement at the smallest time interval is possible.
[0035]
To improve the accuracy, it is necessary to reduce the required accuracy ε, and to increase the polling time interval in inverse proportion to ε. On the other hand, if the reliability p is set to be large (that is, β ′ is large) if the actual fluctuation is set to be large with respect to the measured response delay fluctuation data, the polling time interval is proportional to β ′. Become longer.
[0036]
As described above, the time interval until polling is determined by the polling interval determination unit 1f. Then, a schedule is set by the schedule unit 1g so as to perform polling after the time interval determined in this way.
[0037]
Next, an example of a processing operation of the polling time interval sequential correction system 2 will be described with reference to a flowchart of FIG.
[0038]
FIG. 3 is a flowchart illustrating an example of a processing procedure of the polling time interval sequential correction method according to the present invention.
[0039]
In the procedure of this example, the initial value is used as the polling time interval until the response delay data for a certain period is accumulated after the measurement is started, and the response delay fluctuation is evaluated sequentially after the response delay data for a certain period is accumulated. Then, the next polling time interval is sequentially determined based on this.
[0040]
That is, when the measurement is started (step 301), sets an initial value _{(Q 0} seconds) as the polling time interval (Q) (step 302). Since historical response delay data of X hours worth of yet predetermined in this state is not stored (step 303), performs polling at the time interval of the initial value (Q ₀ seconds) (step 304).
[0041]
The response delay time at that time is accumulated (step 305), and a measured value of, for example, throughput is calculated and displayed based on the data acquired by polling (step 306). Then, (in this case Q after ₀ seconds of the initial value) for the next poll after Q seconds schedule as (step 307).
[0042]
Repeat polling of the initial value Q per ₀ seconds, the response delay data for the past X times is accumulated (step 303), and evaluating the response delay fluctuation on the basis of the response delay data (step 308), the result evaluation , A polling time interval (Q = β · α) from the next time is obtained (step 309).
[0043]
Thereafter, the polling is performed (step 304), the response delay time is accumulated (step 305), the measured value is calculated and displayed (step 306), and the next polling schedule is set to (Q = β · α). (Step 307), the processing from Step 303 is repeated.
[0044]
As described above with reference to FIGS. 1 to 5, in the polling time interval sequential correction system and method and the throughput measuring device of the present example, the polling response delay time is measured, and the i-th polling response delay time is calculated. Ti is sequentially recorded, and the latest response delay fluctuation is calculated based on the data of the response delay time Ti (i = 1... N) from a certain time past to the present time, and under this response delay fluctuation, The polling time interval for satisfying the required measurement accuracy is calculated. Then, based on the polling time interval determined in this way, a measurement such as a throughput at a specific link on the network is executed.
[0045]
As described above, in the present example, the polling interval is not determined in advance as in the related art, but the polling interval is determined according to the polling response delay fluctuation that changes depending on the network conditions.
[0046]
As a result, in the related art, the polling for obtaining the flow counter value is performed using a predetermined constant value, so that the measurement accuracy cannot be suppressed to a constant level. According to the present invention, the polling time interval is determined based on the polling response delay fluctuation at the present time, so that the error of the time interval can be suppressed to a certain level or less, and the measurement error can be kept at the required level or less. Measurement accuracy can be maintained at a constant level regardless of the state of the network.
[0047]
The present invention is not limited to the examples described with reference to FIGS. 1 to 5 and can be variously modified without departing from the gist thereof. For example, in this example, SNMP polling, which is a network management system, has been described as an example, but the present invention is also applicable to a polling function installed in another system.
[0048]
In this example, the throughput measuring device 1 that measures the throughput of the link 23 is used for the description of the measuring device. However, for example, a device that measures the CPU usage rate of the router 21 may be used.
[0049]
In this embodiment, the optical disk is used as a recording medium, but the FD may be used as a recording medium. As for the installation of the program, the program may be downloaded and installed via a network via a communication device.
[0050]
【The invention's effect】
According to the present invention, the polling time interval can be determined as small as possible according to the response delay situation in the network, and the rate of increase per unit time, such as the throughput on a predetermined link by polling at a remote place, can be determined. Measurement accuracy can be improved.
[Brief description of the drawings]
FIG. 1 is a block diagram illustrating a configuration example of a polling time interval sequential correction system according to the present invention and a throughput measurement device including the same.
FIG. 2 is a block diagram illustrating a configuration example of a network to be measured by the throughput measuring device in FIG. 1;
FIG. 3 is a flowchart illustrating an example of a processing procedure of a polling time interval sequential correction method according to the present invention.
FIG. 4 is a histogram showing a distribution example of a difference between response delay times in an IP network.
FIG. 5 is a block diagram illustrating an example of a hardware configuration of a throughput measurement device including the measurement device polling time interval sequential correction system in FIG. 1;
[Explanation of symbols]
1: Throughput measurement device, 1a: Polling packet sending unit, 1b: Response packet (counter value) receiving unit, 1c: Polling response delay time measuring unit, 1d: Response delay data storage memory, 1e: Polling response delay fluctuation calculating unit, 1f: Polling interval determination unit, 1g: Polling schedule unit, 1h: Bandwidth calculation unit, 2: Polling time interval sequential correction system, 20: Communication network (IP network), 21, 22: Router ("node"), 23: Link ("measurement target link"), 24: MIB counter, 51: display device, 52: input device, 53: external storage device, 54: information processing device, 54a: CPU, 54b: main memory, 55: optical disk, 56 : Drive device, 57: communication device.

Claims (7)

First means for sequentially measuring a polling response delay time and recording the result in a storage device;
The response delay time from the present time to a predetermined past time recorded in the storage device is read, and a difference between two adjacent response delay times and a distribution of the difference are calculated based on the read response delay time. A second means for
Based on the distribution of difference calculated by the second means, it can be suppressed to below a pre requested value of the error of the measurements taken in the polling, and a third means for calculating a shortest polling time interval Have
A polling time interval sequential correction system, wherein the polling time interval is sequentially corrected to the polling time interval calculated by the third means . First means for sequentially measuring a polling response delay time and recording the result in a storage device;
The response delay time from the present time to a predetermined past time recorded in the storage device is read, and based on the read response delay time, a difference Xi between two adjacent response delay times and a standard value of the difference Xi. Second means for calculating the deviation σ,
From the standard deviation σ calculated by the second means, the difference Xi, and the reliability p used for setting the magnitude of fluctuation of the polling time interval, Prob ｛−β ′ · σ ≦ Xi ≦ β ′ · σ｝ = P, and using the parameter β ', the standard deviation σ, and the required accuracy ε, which is an error condition of the measurement value obtained in the previously requested polling, until the next polling. Third means for calculating the time interval S according to S = ｛(2 / ε) +1｝ β ′ · σ,
A polling time interval sequential correction system, wherein the polling time interval is sequentially corrected to the polling time interval calculated by the third means. A first step of sequentially measuring a polling response delay time and recording the polling response delay time in a storage device;
The response delay time from the present time to a predetermined past time recorded in the storage device is read, and a difference between two adjacent response delay times and a distribution of the difference are calculated based on the read response delay time. A second step to
A third step of calculating a shortest polling time interval capable of suppressing an error of a measurement value obtained by polling to be equal to or less than a previously required value based on the distribution of the difference calculated in the second step; Has,
A polling time interval sequential correction method, wherein the polling time interval is sequentially corrected to the polling time interval calculated in the third step. A first step of sequentially measuring a polling response delay time and recording the polling response delay time in a storage device;
The response delay time from the present time to a predetermined past time recorded in the storage device is read, and based on the read response delay time, a difference Xi between two adjacent response delay times and a standard value of the difference Xi. A second step of calculating the deviation σ,
From the standard deviation σ calculated in the second step, the difference Xi, and the reliability p used for setting the magnitude of fluctuation of the polling time interval, Prob ｛−β ′ · σ ≦ Xi ≦ β ′ · σ｝ = P, and using the parameter β ', the standard deviation σ, and the required accuracy ε, which is an error condition of the measurement value obtained in the previously requested polling, until the next polling. A third step of calculating the time interval S by S = ｛(2 / ε) +1｝ β ′ · σ,
A polling time interval sequential correction method, wherein the polling time interval is sequentially corrected to the polling time interval calculated in the third step. A computer-readable recording medium having recorded thereon a program for causing a computer to execute each step in the polling time interval sequential correction method according to claim 3. First means for sequentially measuring a polling response delay time and recording the result in a storage device;
The response delay time from the present time to a predetermined past time recorded in the storage device is read, and a difference between two adjacent response delay times and a distribution of the difference are calculated based on the read response delay time. A second means for
A third means for calculating a shortest polling time interval, which can suppress an error of a measured value obtained by polling to be equal to or less than a previously requested value, based on a distribution of the difference calculated by the second means;
A fourth means for requesting notification of data for calculating a measurement value to a device to be measured at the polling time interval calculated by the third means;
Fifth means for obtaining the measured value based on the measured value calculation data notified from the device to be measured in response to the request from the fourth means. apparatus. First means for sequentially measuring a polling response delay time and recording the result in a storage device;
The response delay time from the present time to a predetermined past time recorded in the storage device is read, and based on the read response delay time, a difference Xi between two adjacent response delay times and a standard value of the difference Xi. Second means for calculating the deviation σ,
From the standard deviation σ calculated by the second means, the difference Xi, and the reliability p used for setting the magnitude of fluctuation of the polling time interval, Prob ｛−β ′ · σ ≦ Xi ≦ β ′ · σ｝ = P, and using the parameter β ', the standard deviation σ, and the required accuracy ε, which is an error condition of the measurement value obtained in the previously requested polling, until the next polling. Third means for calculating the time interval S by S = ｛(2 / ε) +1｝ β ′ · σ;
A fourth means for requesting notification of data for calculating a measurement value to a device to be measured at the polling time interval calculated by the third means;
Fifth means for obtaining the measured value based on the measured value calculation data notified from the device to be measured in response to the request from the fourth means. apparatus.

Images