JP4750732B2 - Path capacity increase / decrease judgment method and path communication device - Google Patents

Description

  The present invention communicates IP packets, MPLS packets, Ethernet (registered trademark) frames and the like between communication devices such as IP (Internet Protocol) routers via TDM paths, MPLS paths, wavelength paths (optical paths), and the like. Used for path network In particular, the present invention relates to a technique for increasing / decreasing path capacity in accordance with the traffic volume transmitted by the path.

  Data communication traffic including the Internet is increasing rapidly. A communication network that supports this requires a large-capacity network device and a large-capacity communication system in order to process a large amount of traffic and perform communication.

  A communication system such as SONET or SDH used as a communication path in a network such as IP or Ethernet is a system in which a path is fixedly set between the start point and the end point of the communication path. A path is set by designating a time slot multiplexed time slot. Once a path is set, a certain capacity is always secured in the network.

  In recent years, there is a technique for providing a large-capacity communication path using wavelength division multiplexing (WDM) technology, but the capacity of the communication path is fixed as in SDH and SONET. In Ethernet, a method of using a plurality of links as a single virtual link has been put into practical use by using a technique standardized by IEEE 802.3ad called ring aggregation.

  On the other hand, in a lower layer, a technology capable of providing a single virtual large-capacity path by bundling a plurality of SONET and SDH paths using a technique called virtual concatenation has been developed. Both ring aggregation and virtual concatenation are technologies that combine a plurality of paths or physical links to provide a large-capacity communication path, and the set bandwidth is fixed.

  In the current network using the above SDH / SONET or WDM as a communication path of a packet network such as the Internet using an IP router or an Ethernet device, the capacity of the communication path is fixed as described above. For this reason, it is impossible to follow fluctuations in traffic over time. Therefore, at present, it is necessary that a communication operator predicts the maximum traffic amount in advance and prepares a transmission path having a capacity capable of transmitting the maximum traffic with a margin.

  In this case, as in the case of the Internet, when the traffic fluctuation is large and it corresponds to the increased traffic in recent years, a communication path with a much larger capacity is ensured in normal times compared to the actual traffic volume. And the capacity in the network cannot be used efficiently.

  Internet traffic is traffic in which traffic generated from a large number of users is concentrated, and has several characteristics. The first feature is that there is a gradual change in traffic volume depending on the number of users using the network and the frequency of use, such as the time fluctuation of one day. The second feature is that there is a short-term random fluctuation called microburst caused by bursty traffic that occurs when a user actually uses the network.

  As a factor of short-term fluctuation, the form of data transfer from the WWW site to the user when browsing the WWW (World Wide Web) site on the Internet or the occurrence of file exchange traffic is bursty. As a result, a short-term and random fluctuation occurs as a characteristic of the upper layer traffic.

  A third feature is traffic fluctuation caused by a failure of an IP router constituting the Internet. This occurs when the routing table is changed due to a failure of the IP router or the communication path, and the packet transfer destination is changed.

  Traffic fluctuations due to failures occur suddenly and cannot be predicted. In addition, when a plurality of autonomous networks are connected to each other like the Internet, there may be a case where other networks fail to cause traffic fluctuations, and furthermore, the failure may not be detected. .

  FIG. 11 and FIG. 12 show examples of traffic fluctuations caused by long-term fluctuations of Internet traffic and sudden failures, respectively. In the example of FIG. 11, the traffic is lowest in the morning and highest near midnight. In the example of FIG. 12, it can be seen that the traffic is rapidly increasing immediately after the failure occurs.

  Non-Patent Document 1 describes a variable capacity technique considering long-term fluctuation traffic and short-term fluctuation of traffic. Non-Patent Document 1 observes traffic, approximates the amount of traffic over a certain period of time, predicts the amount of traffic due to long-term time fluctuations, and calculates short-term and random time fluctuations obtained from measurement data In addition, there is described a method for realizing a variable capacity link that determines whether the path capacity is increased or decreased by combining the traffic amount considering the predicted long-term fluctuation and the short-term traffic fluctuation.

  In Non-Patent Document 2, the Kalman filter is used to decompose traffic into a trend component that shows a linear increase / decrease trend and a periodic variation component, and based on this, band acceptance prediction is performed, thereby providing high follow-up to periodic variations. Capacitive capacity control is realized.

Okazaki Yasutaka, Watanabe Atsushi, Takahashi Tetsuo, Tsukishima Yukio, “Proposal of Variable Capacity Optical Path System Considering Burstness of User Traffic”, IEICE Tech. 35-38, May 2003. IEICE Technical Report SSE95-120, Shioda et al., “Equipment Design Method in Self-Sizing Network”

  In both Non-Patent Documents 1 and 2, the change function is known after observing periodic fluctuations in the traffic volume, and the predicted value obtained therefrom is used for capacity control.

  However, in Internet traffic, it is difficult to model the occurrence pattern of traffic, so it is difficult to make known the change function of the long-term traffic tendency. In addition, there may be a case where the periodic tendency is not necessarily extracted accurately because of being buried in random short-term fluctuations, and accurate capacity control becomes difficult. Furthermore, there is a problem that the technique of Non-Patent Document 1 or 2 cannot cope with non-periodic fluctuations.

  In view of this point, the present invention extracts the behavior of the non-random component from the finite traffic observation data even when the component (non-random component) that exhibits the non-random behavior in the traffic fluctuation is unknown, and the capacity control based on the extracted behavior. The goal is to minimize traffic.

  The present invention extracts parameters representing the characteristics of random components from finite traffic observation data, and sequentially optimizes digital filter operations based on the parameters, thereby obtaining non-random traffic components as filter output data. This is a characteristic path capacity increase / decrease determination method. The path capacity increase / decrease determination is performed based on the non-random component obtained in this way.

  That is, the present invention provides a traffic observation step in which a path communication device that provides a communication path by a path observes the traffic amount transmitted by the path as discrete-time traffic by sampling, and the traffic amount observed by the traffic observation step. A path capacity increase determination step for determining a path capacity increase based on the path capacity decrease determination step for determining a path capacity decrease based on the traffic volume observed in the traffic observation step, the path capacity increase determination step, and the path By executing the path capacity change instruction step that instructs to increase or decrease the path capacity based on the determination result of the capacity decrease determination step, the path capacity in the path communication network is determined according to the traffic volume transmitted by the path. This is a path capacity increase / decrease determination method to be changed.

  Here, a feature of the present invention is that in the path capacity increase determination step and the path capacity decrease determination step, the actual measurement data is obtained by performing digital filter operation on the basis of past actual measurement data in the traffic observation step. Even when random traffic fluctuations are present, the tendency to increase or decrease the traffic is extracted by removing the random component, and the random component is handled as a margin.

  For example, when the traffic is assumed to be traffic composed of a random component and a non-random component, the path capacity increase determination step and the path capacity decrease determination step are performed based on the correlation between the traffic and past measured data. The first step of extracting the characteristics of the random component, the second step of calculating the non-random component of traffic by applying a digital filter operation to the actual measurement data, the actual measurement data, and the second step A third step of extracting the characteristics of the random component of the traffic reflecting the influence of the digital filter operation based on the non-random component of the traffic, and the random component of the traffic obtained from the first and third steps, respectively. Based on the characteristics of the fourth to correct the calculation of the filter operation Step 5 and the fifth step of determining the latest digital filter operation calculation to be performed on the next measured data based on the correction value of the filter operation calculation obtained in the fourth step. Each time the actual measurement data is acquired, the traffic increasing tendency or decreasing tendency is sequentially extracted by repeatedly executing the first to fifth steps.

  For example, assuming that the past j actually measured data are x (n−1), x (n−2),..., X (n−j) in order, the path capacity increase determination step and the path capacity decrease determination. The digital filter operation in the step is performed from two actually measured data continuously in time among the j actually measured data.

に基づいてランダム成分の特徴σ_s（ｎ）を算出する第六のステップと、最新のフィルタ出力データをｙ（ｎ）とし、

A sixth step of calculating the characteristic σ _s (n) of the random component based on, and the latest filter output data as y (n),

で表されるフィルタ操作により、ｙ（ｎ−ｊ）〜ｙ（ｎ）を求める第七のステップと、フィルタ出力データｙ（ｌ）と、実測データｘ（ｌ）とを基に、

On the basis of the seventh step for obtaining y (n−j) to y (n) by the filter operation represented by the following equation, the filter output data y (l), and the actual measurement data x (l):

によりフィルタ操作の影響を反映したトラヒックのランダム成分の特徴ｐ_iを算出する第八のステップと、前記第六および第八のステップのそれぞれで算出したランダム成分の特徴σ_s、ｐ_iを基に、
Δｗ_i（ｎ）＝α（ｐ_i（ｎ）＋σ_s ²（ｎ）δ_0i）
（但し、ｋは任意の整数、ｉ＝０、１、…、ｋ、δはクロネッカーデルタ、αは実数）
によりフィルタ係数ｗ₀（ｎ）、ｗ₁（ｎ）、…、ｗ_k（ｎ）の補正値Δｗ₀（ｎ）、Δｗ₁（ｎ）、…、Δｗ_k（ｎ）を求める第九のステップと、補正値を考慮した最新のフィルタを
ｗ_i（ｎ＋１）＝ｗ_i（ｎ）＋Δｗ_i（ｎ） （但し、ｉ＝０、１、…、ｋ）
として構成する第十のステップとを有する。

Based the eighth step of calculating a characteristic p _i of the random component of the traffic that reflects the influence of the filtering operation, characterized sigma _s, p _i of the random component calculated in each of the sixth and eighth steps by ,
Δw _i (n) = α (p _i (n) + σ _s ² (n) δ _0i )
(Where k is an arbitrary integer, i = 0, 1,..., K, δ is a Kronecker delta, α is a real number)
Filter coefficients _{w 0 (n), w 1} (n) by, ..., the correction value [Delta] w of _{w k (n) 0 (n} ), Δw 1 (n), ..., the ninth step of obtaining a Δw _k (n) And the latest filter considering the correction value w _i (n + 1) = w _i (n) + Δw _i (n) (where i = 0, 1,..., K)
And a tenth step.

  Alternatively, assuming that the past j actually measured data are sequentially x (n−1), x (n−2),..., X (n−j), the path capacity increase determination step and the path capacity decrease determination. The digital filter operation in the step is performed from two actually measured data continuously in time among the j actually measured data.

に基づいてランダム成分の特徴σ_s（ｎ）を算出する第十一のステップと、最新のフィルタ出力データをｙ（ｎ）とし、

The eleventh step of calculating the characteristic σ _s (n) of the random component based on, and the latest filter output data as y (n),

で表されるフィルタ操作により、ｙ（ｎ−ｊ）〜ｙ（ｎ）を求める第十二のステップと、フィルタ出力データｙ（ｌ）と、実測データｘ（ｌ）とを基に、

Based on the twelfth step of obtaining y (n−j) to y (n), the filter output data y (l), and the actual measurement data x (l) by the filter operation represented by

によりフィルタ操作の影響を反映したトラヒックのランダム成分の特徴ｐ_iを算出する第十三のステップと、前記第十一および第十三のステップのそれぞれで算出したランダム成分の特徴σ_s、ｐ_iを基に、
Δｗ_i（ｎ）＝αｐ_i（ｎ）
（但し、ｋは任意の整数、ｉ＝０、１、…、ｋ、αは実数）
によりフィルタ係数ｗ₀（ｎ）、ｗ₁（ｎ）、…、ｗ_k（ｎ）の補正値Δｗ₀（ｎ）、Δｗ₁（ｎ）、…、Δｗ_k（ｎ）を求める第十四のステップと、補正値を考慮した最新のフィルタを
ｗ_i（ｎ＋１）＝ｗ_i（ｎ）＋Δｗ_i（ｎ） （但し、ｉ＝０、１、…、ｋ）
として構成する第十五のステップとを有する。

The random component features σ _s , p _i calculated in each of the thirteenth step and the eleventh and thirteenth steps of calculating the traffic random component feature p _i reflecting the influence of the filtering operation Based on
Δw _i (n) = αp _i (n)
(Where k is an arbitrary integer, i = 0, 1,..., K, α are real numbers)
, The filter coefficients _{w 0 (n), w 1} (n) by ..., the correction value [Delta] w of _{w k (n) 0 (n} ), Δw 1 (n), ..., 14th seeking Δw _k (n) The latest filter considering the step and the correction value is represented by w _i (n + 1) = w _i (n) + Δw _i (n) (where i = 0, 1,..., K)
And the fifteenth step.

  For example, the filter shape in the seventh step of the digital filter operation is

で表されるものとし、前記第九のステップが

And the ninth step is

である。

It is.

  Alternatively, the filter shape in the twelfth step of the digital filter operation is

で表されるものとし、前記第十四のステップが

It is assumed that the fourteenth step is

である。

It is.

  Alternatively, the filter shape in the seventh or twelfth step of the digital filter operation is

であることもできる。

It can also be.

  Alternatively, the filter shape in the seventh or twelfth step of the digital filter operation is

であることもできる。

It can also be.

For example, in the path capacity increase determination step and the path capacity decrease determination step, when the capacity of the path increase / decrease unit is L and the maximum capacity of the link is M, M = bL (b is a natural number), When the actually measured data in the traffic observation step for the path a is the sum of the traffic passing through the path a,
y (n) + p · σ _d (n)> aL (a is a natural number less than b, p is a positive real number)
In the case of
y (n) + q · σ _d (n) <(a−1) L (q is a positive real number greater than or equal to p)
In case of, reduce the path.

Alternatively, in the path capacity increase determination step and the path capacity decrease determination step, when the capacity of the path increase / decrease unit is L and the maximum capacity of the link is M, M = bL (b is a natural number), When the actual measurement data in the traffic observation step is the traffic amount for each path,
When the number of existing paths is a,
y (n) + p · σ _d (n)> L (p is a positive real number)
In the case of, increase the number of paths, and in all the a paths,
y (n) + q · σ _d (n) <L (q is a positive real number greater than or equal to p)
In the case of, it is possible to reduce the path.

  The present invention can also be viewed from the viewpoint of a path communication device. That is, the present invention samples traffic volume transmitted by a path and observes it as discrete-time traffic, and non-random traffic based on the traffic volume observed by the traffic monitoring means and past measured data. Filter output data representing a component, digital filter means for calculating a standard deviation of the traffic amount with respect to the filter output data, and path capacity increase determination means for sequentially extracting an increase tendency of traffic based on the filter output data and the standard deviation And a path capacity decrease determination means for sequentially extracting a traffic decrease tendency based on the filter output data and the standard deviation, and a path capacity increase / decrease based on a determination result of the path capacity increase determination means or the path capacity decrease determination means. Instruction to change path capacity The path communication device and a stage.

  For example, the digital filter means extracts a feature of a random component of traffic from the correlation between the traffic amount observed by the traffic observation means and past measured data, and performs a digital filter operation on the traffic amount. Calculates filter output data representing non-random traffic components, extracts the random component features from the filter output data and the traffic volume, and creates a digital filter based on the random component features extracted by these two different methods. The calculation value of the calculation in the operation is calculated, and the operation for determining the calculation of the digital filter operation to be performed on the next actual measurement data based on the correction value is repeatedly executed every time the actual measurement data is acquired from the traffic observation means. And the traffic volume for the filter output data Comprising means for calculating a standard deviation.

  For example, assuming that the past j actual measurement data are sequentially x (n−1), x (n−2),..., X (n−j), the digital filter means performs the digital filter operation. As a means of the above, from two actually measured data continuously in time among the j actually measured data

に基づいてランダム成分の特徴σ_s（ｎ）を算出する手段と、最新のフィルタ出力データをｙ（ｎ）とし、

A means for calculating the characteristic σ _s (n) of the random component based on the above, and the latest filter output data as y (n),

で表されるフィルタ操作により、ｙ（ｎ−ｊ）〜ｙ（ｎ）を求める手段と、フィルタ出力データｙ（ｌ）と、実測データｘ（ｌ）とを基に、

Based on the means for obtaining y (n−j) to y (n), the filter output data y (l), and the actual measurement data x (l) by the filter operation represented by

によりフィルタ操作の影響を反映したトラヒックのランダム成分の特徴ｐ_iを算出する手段と、前記特徴σ_s（ｎ）を算出する手段および特徴ｐ_iを算出する手段のそれぞれで算出したランダム成分の特徴σ_s、ｐ_iを基に、
Δｗ_i（ｎ）＝α（ｐ_i（ｎ）＋σ_s ²（ｎ）δ_0i）
（但し、ｋは任意の整数、ｉ＝０、１、…、ｋ、δはクロネッカーデルタ、αは実数）
によりフィルタ係数ｗ₀（ｎ）、ｗ₁（ｎ）、…、ｗ_k（ｎ）の補正値Δｗ₀（ｎ）、Δｗ₁（ｎ）、…、Δｗ_k（ｎ）を求める手段と、補正値を考慮した最新のフィルタを
ｗ_i（ｎ＋１）＝ｗ_i（ｎ）＋Δｗ_i（ｎ） （但し、ｉ＝０、１、…、ｋ）
として構成する手段とを含む。

The feature of the random component calculated by the means for calculating the feature p _i of the random component of the traffic reflecting the influence of the filter operation, the means for calculating the feature σ _s (n), and the means for calculating the feature p _i. Based on σ _s and p _i ,
Δw _i (n) = α (p _i (n) + σ _s ² (n) δ _0i )
(Where k is an arbitrary integer, i = 0, 1,..., K, δ is a Kronecker delta, α is a real number)
Filter coefficients _{w 0 (n), w 1} (n) by, ..., the correction value [Delta] w of _{w k (n) 0 (n} ), Δw 1 (n), ..., and means for determining the [Delta] w _k (n), corrected The latest filter in consideration of the value w _i (n + 1) = w _i (n) + Δw _i (n) (where i = 0, 1,..., K)
And means for configuring as follows.

  Alternatively, assuming that the past j actual measurement data are sequentially x (n−1), x (n−2),..., X (n−j), the digital filter means performs the digital filter operation. As a means of the above, from two actually measured data continuously in time among the j actually measured data

に基づいてランダム成分の特徴σ_s（ｎ）を算出する手段と、最新のフィルタ出力データをｙ（ｎ）とし、

A means for calculating the characteristic σ _s (n) of the random component based on the above, and the latest filter output data as y (n),

で表されるフィルタ操作により、ｙ（ｎ−ｊ）〜ｙ（ｎ）を求める手段と、フィルタ出力データｙ（ｌ）と、実測データｘ（ｌ）とを基に、

Based on the means for obtaining y (n−j) to y (n), the filter output data y (l), and the actual measurement data x (l) by the filter operation represented by

によりフィルタ操作の影響を反映したトラヒックのランダム成分の特徴ｐ_iを算出する手段と、前記特徴σ_s（ｎ）を算出する手段および特徴ｐ_iを算出する手段のそれぞれで算出したランダム成分の特徴σ_s、ｐ_iを基に、
Δｗ_i（ｎ）＝αｐ_i
（但し、ｋは任意の整数、ｉ＝０、１、…、ｋ、αは実数）
によりフィルタ係数ｗ₀（ｎ）、ｗ₁（ｎ）、…、ｗ_k（ｎ）の補正値Δｗ₀（ｎ）、Δｗ₁（ｎ）、…、Δｗ_k（ｎ）を求める手段と、補正値を考慮した最新のフィルタを
ｗ_i（ｎ＋１）＝ｗ_i（ｎ）＋Δｗ_i（ｎ） （但し、ｉ＝０、１、…、ｋ）
として構成する手段とを含む。

The feature of the random component calculated by the means for calculating the feature p _i of the random component of the traffic reflecting the influence of the filter operation, the means for calculating the feature σ _s (n), and the means for calculating the feature p _i. Based on σ _s and p _i ,
Δw _i (n) = αp _i
(Where k is an arbitrary integer, i = 0, 1,..., K, α are real numbers)
Filter coefficients _{w 0 (n), w 1} (n) by, ..., the correction value [Delta] w of _{w k (n) 0 (n} ), Δw 1 (n), ..., and means for determining the [Delta] w _k (n), corrected The latest filter in consideration of the value w _i (n + 1) = w _i (n) + Δw _i (n) (where i = 0, 1,..., K)
And means for configuring as follows.

  For example, the filter shape in the means for obtaining y (n−j) to y (n) is

で表されるものとし、前記フィルタ係数ｗ₀（ｎ）、ｗ₁（ｎ）、…、ｗ_k（ｎ）の補正値Δｗ₀（ｎ）、Δｗ₁（ｎ）、…、Δｗ_k（ｎ）を求める手段が

And in those represented, the filter coefficients _{w 0 (n), w 1} (n), ..., the correction value [Delta] w of _{w k (n) 0 (n} ), Δw 1 (n), ..., Δw k (n )

である。

It is.

  Alternatively, the filter shape in the means for obtaining y (n−j) to y (n) is

で表されるものとし、前記フィルタ係数ｗ₀（ｎ）、ｗ₁（ｎ）、…、ｗ_k（ｎ）の補正値Δｗ₀（ｎ）、Δｗ₁（ｎ）、…、Δｗ_k（ｎ）を求める手段が

And in those represented, the filter coefficients _{w 0 (n), w 1} (n), ..., the correction value [Delta] w of _{w k (n) 0 (n} ), Δw 1 (n), ..., Δw k (n )

である。

It is.

  Alternatively, the filter shape in the means for obtaining y (n−j) to y (n) is

であることもできる。

It can also be.

  Alternatively, the filter shape in the means for obtaining y (n−j) to y (n) is

であることもできる。

It can also be.

For example, the path capacity increase determination means and the path capacity decrease determination means have M = bL (b is a natural number) where L is the capacity of the path increase / decrease unit and M is the maximum capacity of the link, When the measured data in the traffic observing means at the time of path a is the sum of traffic passing through path a,
y (n) + p · σ _d (n)> aL (a is a natural number less than b, p is a positive real number)
In the case of
y (n) + q · σ _d (n) <(a−1) L (q is a positive real number greater than or equal to p)
In this case, means for reducing the path is included.

Alternatively, the path capacity increase determination unit and the path capacity decrease determination unit have M = bL (b is a natural number), where L is the capacity of the path increase / decrease unit and M is the maximum capacity of the link, When the measured data in the traffic observation means is the traffic amount for each path,
When the number of existing paths is a,
y (n) + p · σ _d (n)> L (p is a positive real number)
In the case of, increase the number of paths, and in all the a paths,
y (n) + q · σ _d (n) <L (q is a positive real number greater than or equal to p)
In this case, a means for reducing the path may be included.

  The present invention can also be viewed from the viewpoint of a program. That is, the present invention is a program that, when installed in a general-purpose information processing apparatus, causes the general-purpose information processing apparatus to realize functions corresponding to the path communication apparatus of the present invention.

  By recording the program of the present invention on a recording medium, the general-purpose information processing apparatus can install the program of the present invention using this recording medium. Alternatively, the program of the present invention can be directly installed on the general-purpose information processing apparatus via a network from a server that holds the program of the present invention.

  Thereby, a function corresponding to the path communication device of the present invention can be realized using a general-purpose information processing device.

  The program of the present invention includes not only a program that can be directly executed by a general-purpose information processing apparatus but also a program that can be executed by installing it on a hard disk or the like. Also included are those that are compressed or encrypted.

  According to the present invention, capacity control can be performed based on finite traffic observation data even when non-random components in traffic fluctuation are unknown, and traffic crossing can be minimized.

(First Example)
A first embodiment of the present invention will be described with reference to FIGS. FIG. 1 is a functional block configuration diagram for extracting an increase / decrease tendency of traffic. FIG. 2 is a block diagram of the path communication apparatus.

In the present embodiment, the feature of the random component r (n) and the feature of the non-random component s (n) are sequentially extracted from the actually measured data x (n) by digital filter operation. It is an Example which shows a mode that a decreasing tendency is extracted. Here, the traffic is composed of a random component r (n) and a non-random component s (n). That is,
x (n) = s (n) + r (n)
Assume that

  A path communication device that provides a communication path by a path has a block configuration as shown in FIG. 2 and samples the traffic amount x (n) transmitted by the path in the traffic observation unit 1 that is responsible for the traffic observation step. Observe as discrete-time traffic (“1” in FIG. 1).

  Next, in the digital filter unit 2 responsible for the digital filter step, from the correlation between the traffic amount x (n) observed in the traffic observation step and the past actual measurement data held in the actual measurement data holding unit 6, A feature of a random component r (n) of traffic is extracted (“2” in FIG. 1).

  Next, the digital filter unit 2 calculates filter output data y (n) representing a non-random component of traffic by performing a digital filter operation on x (n) (“3” in FIG. 1), where The characteristics of the random component r (n) are extracted from y (n) and x (n) obtained in step (4 in FIG. 1).

  Next, the digital filter unit 2 calculates a correction value for calculation in the digital filter operation based on the characteristics of the random components extracted by these two different methods (“5” in FIG. 1). This correction value is held in the filter coefficient holding unit 7 or the parameter holding unit 8.

Furthermore, the digital filter unit 2 determines the operation of the digital filter operation to be performed on the next actual measurement data x (n + 1) based on the stored correction value (“6” in FIG. 1). The digital filter unit 2 repeatedly executes this operation every time the actual measurement data is acquired from the traffic observation unit 1. Further, the standard deviation σ _d (n) of the traffic amount x (n) with respect to the filter output data y (n) is calculated.

The path capacity increase determination unit 3 and the path capacity decrease determination unit 4 are based on the standard deviation σ _d (n) of the traffic amount x (n) with respect to the filter output data y (n) and the filter output data y (n). The increase tendency and the decrease tendency are sequentially extracted, and the path capacity change instruction unit 5 realizes path capacity increase / decrease based only on the behavior of the non-random components.

Next, the operation of the digital filter unit 2 will be described with reference to FIGS. The digital filter unit 2 not only calculates the filter output by the digital filter operation but also corrects the filter coefficient w _i (n).

First, a feature σ _s (n) of a random component of traffic is extracted from the actually measured data x (n). Next, a feature p _i (n) of a random component of traffic reflecting the influence of the filter operation is extracted from the filter output data y (n) and the actually measured data x (n). Next, a correction value Δw _i (n) for the filter coefficient is calculated from the features σ _s (n) and p _i (n) of the random components extracted by the respective methods. The correction value Δw _i (n) is held in the filter coefficient holding unit 7. Next, the latest filter coefficient w _i (n + 1) is determined based on the correction value.

  The filter operation is different between FIG. 3 and FIG.

であり、図４では

And in FIG.

である。

It is.

Here, in the method shown in FIG. 3, the random component is removed only for traffic in which w ₀ satisfying σ _s ² −σ _d ² = 0 is sufficiently smaller than s ² / σ _s ^2. It has the feature that it can. That is, the random component cannot always be removed from traffic having any characteristics. However, for traffic composed of a large number of users, such as traffic exchanged in an Internet service provider (ISP) backbone network, the above condition for w ₀ is satisfied, so such traffic This is effective for removing random components.

  On the other hand, the method shown in FIG. 4 has a feature that the random component can be removed regardless of the feature of the traffic. However, compared with the method of FIG. 3, it has a feature that it is inferior in convergence to a value that realizes removal of random components in the adjustment of the filter coefficient.

  Here, although the digital filter operation is a general filter operation, the filter output data y (n) may be calculated by an operation using a moving average filter, a low-pass filter, or a wavelet transform described later.

(Second embodiment)
A second embodiment of the present invention will be described with reference to FIGS. This embodiment is a case where the shape of the filter in the first embodiment is represented by one parameter T (n), particularly when it is a moving average filter, that is, the filter output y (n) is

と表される場合の実施例である。

It is an Example in the case of being represented.

  By sequentially correcting the past data number T (n) used for calculating the average in the moving average filter when the traffic that is subject to path capacity increase / decrease determination has a non-random component in the low frequency component, Non-random components can be separated by filtering.

  In addition, when there is only one parameter to be corrected in this way, the amount of calculation is reduced compared to the case where many parameters are sequentially corrected as in the first embodiment, and correction in a shorter time is realized. it can. The processing performed in each functional unit of the path communication apparatus is the same as that in the first embodiment except for the digital filter unit 2 '.

Below, operation | movement of digital filter part 2 'is demonstrated. First, the characteristic σ _s (n) of the random component of traffic is calculated from the actually measured data x (n).

Next, the standard deviation σ _d (n) of the actual measurement data x (n) with respect to the filter output data y (n) is calculated. In the present embodiment, this standard deviation is a feature of a random component of traffic that reflects the influence of the filter operation.

Next, a correction value ΔT (n) for the parameter T (n) is calculated from the features σ _s (n) and σ _d (n) of the random components calculated by the respective methods. This correction value ΔT (n) is held in the parameter holding unit 8. Next, the latest parameter T (n + 1) is determined based on the correction value.

  The filter operation is different between FIG. 5 and FIG.

であり、図６では

And in FIG.

である。

It is.

  The effect due to the difference in filter operation in FIGS. 5 and 6 is the same as the effect due to the difference in filter operation in FIGS. 3 and 4.

(Third embodiment)
A third embodiment of the present invention will be described with reference to FIGS. This embodiment is a case where the shape of the filter in the first embodiment is represented by one parameter T (n), particularly when it is a low-pass filter, that is, the filter output y (n) is

と表される場合の実施例である。

It is an Example in the case of being represented.

  When the traffic that is subject to path capacity increase / decrease judgment has a non-random component in the low-frequency component, filter the non-random component by sequentially correcting the cutoff frequency 1 / T (n) in the low-pass filter. Can be separated.

  In addition, when there is only one traffic to be corrected in this way, the amount of calculation is reduced compared to the case where many parameters are sequentially corrected as in the first embodiment, and correction in a shorter time is realized. it can.

  The same effect can be obtained in the second embodiment, but in the second embodiment, the number of past data necessary for calculating the average changes each time. You need to retain the data.

  On the other hand, the filter operation of FIG.

は無限級数であるため、一般に実現できない。そこで、Ｔ（ｎ）≫１として、

Since is an infinite series, it cannot generally be realized. Therefore, T (n) >> 1

を、上記フィルタ操作の代わりとする。

Is used instead of the above filter operation.

  At this time, the past data necessary for realizing this filter operation is one of y (n−1). Thus, unlike the second embodiment, this embodiment has an advantage that it is not necessary to store a lot of past data for the filter operation. The processing performed by each functional unit of the path communication device is the same as that of the first embodiment except for the digital filter unit 2 ″.

Hereinafter, the operation of the digital filter unit 2 ″ will be described. First, the characteristic σ _s (n) of the random component of traffic is calculated from the actually measured data x (n). Next, the filter output data y (n) The standard deviation σ _d (n) of the actually measured data x (n) is calculated, and in this embodiment, this standard deviation is a feature of the random component of traffic reflecting the influence of the filter operation.

Next, a correction value ΔT (n) for the parameter T (n) is calculated from the features σ _s (n) and σ _d (n) of the random components calculated by the respective methods. This correction value ΔT (n) is held in the parameter holding unit 8. Next, the latest parameter T (n + 1) is determined based on the correction value.

  In the example of FIG. 8, the filter operation is

の例である。

It is an example.

  The effect due to the difference in filter operation in FIGS. 7 and 8 is the same as the effect due to the difference in filter operation in FIGS.

(Fourth embodiment)
A fourth embodiment of the present invention will be described with reference to FIG. In the present embodiment, the path capacity increase determination unit 3 and the path capacity decrease when the actually measured data in the traffic observation unit 1 is the sum of traffic passing through the path a when the number of paths is a, such as the TDM path. 3 is an example showing the operation of the determination unit 4.

The increase / decrease of the path capacity is calculated by the digital filter unit 2 as a filter output y (n) representing a non-random component of traffic, and a standard deviation σ _d (n) of measured data x (n) with respect to y (n). Is performed by the path capacity increase determination unit 3 (S1) and the path capacity decrease determination unit 4 (S2).

  Hereinafter, the path capacity increase / decrease unit is L, the maximum link capacity is M, and M = bL (b is a natural number). A is a natural number less than b.

First, in the path capacity increase determination unit 3, y (n) and σ _d (n) are
y (n) + pσ _d (n)> aL
Is checked (S1). If the above expression is satisfied, an instruction to increase the path capacity is issued to the path capacity change instruction unit 5 (S3), and the operation for y (n) and σ _d (n) is terminated. If the above equation is not satisfied (S1), then, at the path capacity decrease determination unit 4,
y (n) + qσ _d (n) <aL
Is checked (S2). If the above equation is satisfied (S2), an instruction to decrease the path capacity is issued to the path capacity change instruction unit 5 (S4), and the operation for y (n) and σ _d (n) is terminated. If the above equation is not satisfied (S2), the path capacity change instruction unit 5 is not instructed to change the path capacity (S5), and the operation for y (n) and σ _d (n) is terminated. Here, p and q are arbitrary real numbers.

  It is also conceivable that y (n) used for increase determination and y (n) used for decrease determination are calculated by digital filter processing using different filters. As an example, y (n) used for increase determination is calculated by a digital filter process using a low-pass filter having a variable parameter T (n), and y (n) used for decrease determination is a digital filter using a low-pass filter having a fixed parameter T. For example, it is calculated by processing.

(Fifth embodiment)
A fifth embodiment of the present invention will be described with reference to FIG. In this embodiment, the operation of the path capacity increase determination unit 3 and the path capacity decrease determination unit 4 when the actual measurement data in the traffic observation unit 1 is the traffic amount for each path, as in the case of the wavelength path, is shown. It is.

The increase / decrease of the path capacity is calculated by the digital filter unit 2 as a filter output y (n) representing a non-random component of traffic, and a standard deviation σ _d (n) of measured data x (n) with respect to y (n). Based on the above, it is performed by the path capacity increase determination unit 3 (S10) and the path capacity decrease determination unit 4 (S11). Hereinafter, the path capacity increase / decrease unit is L, the maximum link capacity is M, and M = bL (b is a natural number). A is a natural number less than b.

First, in the path capacity increase determination unit 3, y (n), σ _d (n)
y (n) + pσ _d (n)> L
Is checked (S10). Here, if even one path satisfies the above equation, the path capacity change instruction unit 5 is instructed to increase the path capacity (S12), y (n), σ _d (n) End the operation. If there is no path satisfying the above equation (S12), the operation proceeds to the operation of the path capacity decrease determination unit 4.

In the path capacity decrease determination unit 4, for all a paths,
y (n) + qσ _d (n) <L
Is checked (S11). Here, if all the paths satisfy the above equation (S11), the path capacity change instruction unit 5 is instructed to reduce the path capacity (S13), and the operation for y (n) and σ _d (n) is performed. finish. If there is even one path that does not satisfy the above equation, the path capacity change instruction unit 5 is not instructed to change the path capacity (S14), and the operation for y (n), σ _d (n) is performed. finish. Here, p and q are arbitrary real numbers.

  It is also conceivable that y (n) used for increase determination and y (n) used for decrease determination are calculated by digital filter processing using different filters. As an example, y (n) used for increase determination is calculated by a digital filter process using a low-pass filter having a variable parameter T (n), and y (n) used for decrease determination is a digital filter using a low-pass filter having a fixed parameter T. For example, it is calculated by processing.

(Sixth embodiment)
The sixth embodiment is a program that, when installed on a general-purpose information processing apparatus, causes the general-purpose information processing apparatus to realize functions corresponding to the path communication apparatus of the present embodiment.

  By recording the program of this embodiment on a recording medium, the general-purpose information processing apparatus can install the program of this embodiment using this recording medium. Alternatively, the program of this embodiment can be installed directly on the general-purpose information processing apparatus via a network from a server holding the program of this embodiment.

  Thereby, a function corresponding to the path communication apparatus of the present embodiment can be realized using a general-purpose information processing apparatus.

  The program of this embodiment includes not only a program that can be directly executed by a general-purpose information processing apparatus but also a program that can be executed by installing it on a hard disk or the like. Also included are those that are compressed or encrypted.

  According to the present invention, traffic crossing can be suppressed to a minimum even in a network such as the Internet where it is difficult to model traffic generation patterns, so that the utility value is great for both network users and operators.

The functional block block diagram for extracting the increase / decrease tendency of traffic. The block block diagram of a path | pass communication apparatus. The figure for demonstrating operation | movement of the digital filter part of a 1st Example (the 1). The figure for demonstrating operation | movement of the digital filter part of a 1st Example (the 2). The figure for demonstrating operation | movement of the digital filter part of a 2nd Example (the 1). The figure for demonstrating operation | movement of the digital filter part of a 2nd Example (the 2). The figure for demonstrating operation | movement of the digital filter part of a 3rd Example (the 1). The figure for demonstrating operation | movement of the digital filter part of a 3rd Example (the 2). The link capacity increase / decrease determination flowchart (TDM path) of the fourth embodiment. The link capacity increase / decrease determination flowchart (wavelength path) of the fifth embodiment. The figure which shows the internet traffic example of a long-term fluctuation | variation (one day). The figure which shows the example of the internet traffic fluctuation | variation by sudden failure etc.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Traffic observation part 2, 2 ', 2 "Digital filter part 3 Path capacity increase judgment part 4 Path capacity decrease judgment part 5 Path capacity change instruction part 6 Actual measurement data holding part 7 Filter coefficient holding part 8 Parameter holding part

Claims (19)

A path communication device that provides a communication path by path,
A traffic observation step of observing the amount of traffic transmitted by the path as discrete-time traffic by sampling;
A path capacity increase determination step for determining a path capacity increase based on the traffic amount observed in the traffic observation step;
A path capacity decrease determination step for performing a path capacity decrease determination based on the traffic amount observed in the traffic observation step;
And a path capacity change instruction step for instructing an increase or decrease in the path capacity based on the determination results of the path capacity increase determination step and the path capacity decrease determination step, and according to the traffic volume transmitted by the path. A path capacity increase / decrease determination method for changing a path capacity in the path communication network,
In the path capacity increase determination step and the path capacity decrease determination step, a digital filter operation is performed based on the past actual measurement data in the traffic observation step, so that even when the actual measurement data shows random traffic variations, random extracting an increasing trend or decreasing trend of traffic in the removal of components, it treats the random component as a margin,
When the traffic is assumed to be traffic composed of a random component and a non-random component, the path capacity increase determination step and the path capacity decrease determination step include:
A first step of extracting features of random components of the traffic from a correlation between the traffic and past measured data;
A second step of calculating a non-random component of traffic by applying a digital filter operation to the measured data;
A third step of extracting the characteristics of the random component of the traffic reflecting the influence of the digital filter operation based on the measured data and the non-random component of the traffic calculated in the second step;
A fourth step of correcting the calculation of the filter operation based on the characteristics of the random components of the traffic obtained from the first and third steps,
A fifth step of determining the latest digital filter operation calculation to be performed on the next actual measurement data based on the correction value of the filter operation calculation obtained in the fourth step;
Run
A path capacity increasing / decreasing method characterized by sequentially extracting an increasing tendency or decreasing tendency of traffic by repeatedly executing the first to fifth steps each time when actual measurement data is acquired . When the past j actual measurement data are traced back to x (n−1), x (n−2),..., X (n−j),
The digital filter operation in the path capacity increase determination step and the path capacity decrease determination step includes:
From two actual measurement data that are continuous in time among j actual measurement data

A sixth step of calculating a random component feature σ _s (n) based on
Let y (n) be the latest filter output data,

A seventh step of obtaining y (n−j) to y (n) by a filter operation represented by:
Based on the filter output data y (l) and the actual measurement data x (l),

An eighth step of calculating a feature p _i of a random component of traffic reflecting the influence of the filter operation by:
Based on the characteristics σ _s and p _i of the random component calculated in each of the sixth and eighth steps,
Δw _i (n) = α (p _i (n) + σ _s ² (n) δ _0i )
(Where k is an arbitrary integer, i = 0, 1,..., K, δ is a Kronecker delta, α is a real number)
Filter coefficients _{w 0 (n), w 1} (n) by, ..., the correction value [Delta] w of _{w k (n) 0 (n} ), Δw 1 (n), ..., the ninth step of obtaining a Δw _k (n) When,
The latest filter in consideration of the correction value is w _i (n + 1) = w _i (n) + Δw _i (n) (where i = 0, 1,..., K)
Path capacity controller method according to claim 1, further comprising a tenth step of configuring a. When the past j actual measurement data are traced back to x (n−1), x (n−2),..., X (n−j),
The digital filter operation in the path capacity increase determination step and the path capacity decrease determination step includes:
From two actual measurement data that are continuous in time among j actual measurement data

An eleventh step of calculating a random component feature σ _s (n) based on:
Let y (n) be the latest filter output data,

A twelfth step of obtaining y (n−j) to y (n) by a filter operation represented by:
Based on the filter output data y (l) and the actual measurement data x (l),

A thirteenth step of calculating a feature p _i of a random component of traffic reflecting the influence of the filter operation by:
Based on the characteristics σ _s and p _i of the random component calculated in each of the eleventh and thirteenth steps,
Δw _i (n) = αp _i (n)
(Where k is an arbitrary integer, i = 0, 1,..., K, α are real numbers)
, The filter coefficients _{w 0 (n), w 1} (n) by ..., the correction value [Delta] w of _{w k (n) 0 (n} ), Δw 1 (n), ..., 14th seeking Δw _k (n) Steps,
The latest filter in consideration of the correction value is w _i (n + 1) = w _i (n) + Δw _i (n) (where i = 0, 1,..., K)
The path capacity increasing / decreasing method according to claim 1, further comprising : The filter shape in the seventh step of the digital filter operation is

And the ninth step is

The path capacity increasing / decreasing method according to claim 2 . The filter shape in the twelfth step of the digital filter operation is

It is assumed that the fourteenth step is

The path capacity increasing / decreasing method according to claim 3 . The filter shape in the seventh or twelfth step of the digital filter operation is

The path capacity increasing / decreasing method according to claim 2 or 3 . The filter shape in the seventh or twelfth step of the digital filter operation is

The path capacity increasing / decreasing method according to claim 2 or 3 . In the path capacity increase determination step and the path capacity decrease determination step, when the capacity of the path increase / decrease unit is L and the maximum capacity of the link is M, M = bL (b is a natural number), and the path a When the actual measurement data in the traffic observation step at the time of the book is the sum of the traffic passing through the path a,
y (n) + p · σ _d (n)> aL (a is a natural number less than b, p is a positive real number)
In the case of
y (n) + q · σ _d (n) <(a−1) L (q is a positive real number greater than or equal to p)
The path capacity increasing / decreasing method according to any one of claims 1 to 7 , wherein a path is reduced in the case of. In the path capacity increase determination step and the path capacity decrease determination step, when the capacity of the path increase / decrease unit is L and the maximum capacity of the link is M, M = bL (b is a natural number), and the traffic When the measured data in the observation step is the traffic volume for each path,
When the number of existing paths is a,
y (n) + p · σ _d (n)> L (p is a positive real number)
In the case of
In all a passes,
y (n) + q · σ _d (n) <L (q is a positive real number greater than or equal to p)
The path capacity increasing / decreasing method according to any one of claims 1 to 7 , wherein a path is reduced in the case of. A traffic observation means for sampling the traffic volume transmitted by the path and observing it as discrete-time traffic;
Filter output data representing non-random components of traffic based on the traffic amount observed by the traffic observation means and past measured data, and digital filter means for calculating a standard deviation of the traffic amount with respect to the filter output data;
Based on the filter output data and the standard deviation, path capacity increase determination means for sequentially extracting an increase tendency of traffic,
Based on the filter output data and the standard deviation, path capacity decrease determination means for sequentially extracting a traffic decrease tendency;
A path capacity change instructing means for increasing or decreasing a path capacity based on a determination result of the path capacity increase determining means or the path capacity decrease determining means ,
The digital filter means includes
A filter representing a non-random component of traffic by extracting features of a random component of traffic from a correlation between a traffic amount observed by the traffic observation means and past measured data, and applying a digital filter operation to the traffic amount. Calculates output data, extracts features of random components from the filter output data and the traffic volume, and calculates correction values for operations in digital filter operations based on the features of random components extracted by these two different methods Then, based on this correction value, the operation of determining the calculation of the digital filter operation to be performed on the next actual measurement data is repeatedly executed every time the actual measurement data is acquired from the traffic observation means, and the filter output data is Includes means to calculate standard deviation of traffic volume
A path communication device characterized by that . When the past j actual measurement data are traced back to x (n−1), x (n−2),..., X (n−j),
The digital filter means, as means for operating the digital filter, from two actually measured data that are temporally continuous from among the j actually measured data.

Means for calculating a random component feature σ _s (n) based on
Let y (n) be the latest filter output data,

Means for obtaining y (n−j) to y (n) by a filter operation represented by:
Based on the filter output data y (l) and the actual measurement data x (l),

Means for calculating a feature p _i of a random component of traffic reflecting the influence of the filter operation by:
Based on the random component features σ _s and p _i calculated by the means for calculating the feature σ _s (n) and the means for calculating the feature p _i , respectively,
Δw _i (n) = α (p _i (n) + σ _s ² (n) δ _0i )
(Where k is an arbitrary integer, i = 0, 1,..., K, δ is a Kronecker delta, α is a real number)
Filter coefficients _{w 0 (n), w 1} (n) by, ..., the correction value [Delta] w of _{w k (n) 0 (n} ), Δw 1 (n), ..., and means for determining the [Delta] w _k (n),
The latest filter in consideration of the correction value is w _i (n + 1) = w _i (n) + Δw _i (n) (where i = 0, 1,..., K)
The path communication device according to claim 10 , comprising: means configured as: When the past j actual measurement data are traced back to x (n−1), x (n−2),..., X (n−j),
The digital filter means, as means for operating the digital filter, from two actually measured data that are temporally continuous from among the j actually measured data.

Means for calculating a random component feature σ _s (n) based on
Let y (n) be the latest filter output data,

Means for obtaining y (n−j) to y (n) by a filter operation represented by:
Based on the filter output data y (l) and the actual measurement data x (l),

Means for calculating a feature p _i of a random component of traffic reflecting the influence of the filter operation by:
Based on the random component features σ _s and p _i calculated by the means for calculating the feature σ _s (n) and the means for calculating the feature p _i , respectively,
Δw _i (n) = αp _i
(Where k is an arbitrary integer, i = 0, 1,..., K, α are real numbers)
Filter coefficients _{w 0 (n), w 1} (n) by, ..., the correction value [Delta] w of _{w k (n) 0 (n} ), Δw 1 (n), ..., and means for determining the [Delta] w _k (n),
The latest filter in consideration of the correction value is w _i (n + 1) = w _i (n) + Δw _i (n) (where i = 0, 1,..., K)
The path communication apparatus according to claim 10 , comprising means configured as: The filter shape in the means for obtaining y (n−j) to y (n) is

And in those represented, the filter coefficients _{w 0 (n), w 1} (n), ..., the correction value [Delta] w of _{w k (n) 0 (n} ), Δw 1 (n), ..., Δw k (n )

The path communication device according to claim 11 . The filter shape in the means for obtaining y (n−j) to y (n) is

And in those represented, the filter coefficients _{w 0 (n), w 1} (n), ..., the correction value [Delta] w of _{w k (n) 0 (n} ), Δw 1 (n), ..., Δw k (n )

The path communication device according to claim 12, wherein The filter shape in the means for obtaining y (n−j) to y (n) is

The path communication device according to claim 11 or 12, wherein The filter shape in the means for obtaining y (n−j) to y (n) is

The path communication device according to claim 11 or 12, wherein The path capacity increase determination means and the path capacity decrease determination means have M = bL (b is a natural number), where L is the capacity of the path increase / decrease unit and M is the maximum capacity of the link, and the path a When the actual measurement data in the traffic observation means at the time of the book is the sum of the traffic passing through the path a,
y (n) + p · σ _d (n)> aL (a is a natural number less than b, p is a positive real number)
In the case of
y (n) + q · σ _d (n) <(a−1) L (q is a positive real number greater than or equal to p)
Path communication apparatus according to any one of claims 10 to 16 including means for degrowth paths in case of. The path capacity increase determination means and the path capacity decrease determination means have M = bL (b is a natural number), where L is the capacity of the path increase / decrease unit and M is the maximum capacity of the link, and the traffic When the measured data in the observation means is the traffic volume for each path,
When the number of existing paths is a,
y (n) + p · σ _d (n)> L (p is a positive real number)
In the case of, increase the number of paths, and in all the a paths,
y (n) + q · σ _d (n) <L (q is a positive real number greater than or equal to p)
Path communication apparatus according to any one of claims 10 to 16 including means for degrowth paths in case of. A program that, when installed in a general-purpose information processing apparatus, causes the general-purpose information processing apparatus to realize a function corresponding to the function of the path communication apparatus according to any one of claims 10 to 18 .

Images