JP4702848B2 - Bandwidth restriction setting method and path setting method in link bandwidth restriction control - Google Patents

Description

  The present invention provides a bandwidth constraint setting method and a path in link bandwidth constraint control in which a bandwidth difference corresponding to a quality class of a path is provided in a link through which the path passes, thereby setting a quality difference corresponding to the quality class for each path. It relates to the setting method.

  In a network that sets paths for different quality classes, a technology is known that gives a difference in the quality of the set paths by limiting the bandwidth that can be used by paths belonging to each quality class on each link. Yes.

Non-Patent Document 1 discloses a technique for empirically obtaining bandwidth constraints in each link based on the average path call volume for each quality class of the path added to each link. The following three types of models have been proposed as methods of giving bandwidth restrictions on each link.
(1) RDM (Russian Doll BC-Model)

  FIG. 4 is a diagram schematically representing the concept of RDM. Here, for the sake of simplicity, description will be made assuming three quality classes. Assuming that the bandwidth assigned to high quality class 0 is CT0, the bandwidth assigned to medium quality class 1 is CT1, and the bandwidth assigned to low quality class 2 is CT2 in the link whose bandwidth is BW To do. In RDM, the following band restriction BC is provided for the band CT allocated to each quality class.

CT0 + CT1 + CT2 ≦ BC0
CT1 + CT2 ≦ BC1
CT2 ≦ BC2
BC0 ≧ BC1 ≧ BC2

If the above bandwidth restriction is not satisfied by setting a new path, the path setting request is lost on the link.
(2) MAM (Maximum Allocation BC-Model)

  FIG. 5 is a diagram schematically representing the concept of MAM. Here, for the sake of simplicity, description will be made assuming three quality classes. Assuming that the bandwidth assigned to high quality class 0 is CT0, the bandwidth assigned to medium quality class 1 is CT1, and the bandwidth assigned to low quality class 2 is CT2 in the link whose bandwidth is BW To do. In the MAM, the following band constraint BC is provided for the band CT allocated to each quality class.

CT0 ≦ BC0
CT1 ≦ BC1
CT2 ≦ BC2

If the above bandwidth restriction is not satisfied by setting a new path, the path setting request is lost on the link.
(3) MAR (Max Allocation with Reservation BC-Model)

  For the sake of simplicity, the MAR concept will be described here assuming three quality classes. In the link where the bandwidth is BW, the assigned bandwidth to the high quality class 0 after setting the path is CT0, the assigned bandwidth to the medium quality class 1 is CT1, and the assigned bandwidth to the low quality class 2 is CT2. And RBW indicates a reserved (free) bandwidth. In the MAR, when the path setting of the quality class i (i = 0, 1, 2) is performed, the following band restrictions are provided for the band CT allocated to each quality class.

CT0 + CT1 + CT2 ≤ BW CT i <BC i
CT0 + CT1 + CT2 ≤ BW-RBW CT i ≥ BC i

  If the above bandwidth restriction is not satisfied by setting a new path of quality class i, the path setting request is lost on the link.

In the conventional link bandwidth constraint control system described above, the bandwidth constraint BC of each link is empirically obtained based on the average path call volume for each quality class applied to each link. Hereinafter, how to determine the bandwidth restriction BC according to the prior art will be described by taking the case of three quality classes as an example.
(1) For RDM

  If the average path call volume of quality class i (i = 0, 1, 2) added to each link is Ai,

BC0 = BW
BC1 = Minimum (Factor1 × BW × (A1 + A2) / (A0 + A1 + A2), BW)
BC2 = Minimum (Factor2 × BW × A2 / (A0 + A1 + A2), BW)

Here, Factor i (i = 1, 2) is determined empirically, for example, Factor 1 = 1.0 and Factor 2 = 1.0.
(2) For MAM

  If the average path call volume of quality class i (i = 0, 1, 2) added to each link is Ai,

BC0 = Minimum (Factor0 × BW × A0 / (A0 + A1 + A2), BW)
BC1 = Minimum (Factor1 × BW × A1 / (A0 + A1 + A2), BW)
BC2 = Minimum (Factor2 × BW × A2 / (A0 + A1 + A2), BW)

Again, Factor i is determined empirically, and for example, Factor 0 = 3.0, Factor 1 = 2.0, and Factor 2 = 1.0 are set.
(3) For MAR

  If the average path call volume of quality class i (i = 0, 1, 2) added to each link is Ai,

BC0 = Minimum (Factor0 × BW × A0 / (A0 + A1 + A2), BW-RBW)
BC1 = Minimum (Factor1 × BW × A1 / (A0 + A1 + A2), BW-RBW)
BC2 = Minimum (Factor2 × BW × A2 / (A0 + A1 + A2), BW-RBW)
RBW = Maximum (1.0, BW × Alpha)

Here, Factor i and Alpha are determined empirically, and for example, Factor 0 = 2.0, Factor 1 = 1.5, Factor 2 = 1.0, and Alpha = 0.01 are set.
J. Ash, "Max Allocation with Reservation Bandwidth Constraints Model for Diffserv-aware MPLS Traffic Engineering & Performance Comparisons," IETF RFC4126, June 2005.

  In the above-described prior art, the bandwidth constraint BC has been determined empirically regardless of the path characteristics, and thus the bandwidth constraint BC cannot be set to an appropriate value.

  An object of the present invention is to solve the above-described problems of the prior art, and link bandwidth constraint that can appropriately and appropriately set and update the bandwidth constraint BC according to the quality class of the path based on the characteristics of the path of each quality class. It is an object to provide a bandwidth restriction setting method and a path setting method in control.

To achieve the above object, the present invention provides the following procedure in a link bandwidth constraint control method for setting a quality difference for each path by providing a bandwidth constraint in accordance with a path quality class for a link of a network. It is characterized by including.
(1) Based on the procedure for measuring the bandwidth used for each path for each quality class, the procedure for obtaining the revenue obtained by setting a path on a link based on the path bandwidth used for each quality class, and the bandwidth used for the path For a path setup request that shows a predetermined relationship between revenue and revenue loss, the procedure for obtaining revenue loss for each quality class, which is damaged by setting a path on the link, And a procedure for setting a bandwidth constraint to be assigned to each quality class so as not to be satisfied.
(2) It is characterized in that it includes a procedure for updating the bandwidth constraint to be assigned to each quality class by periodically repeating each procedure from the measurement of the used bandwidth to the setting of the bandwidth constraint.
(3) It is characterized by setting bandwidth restrictions to be assigned to each quality class so that the bandwidth restriction condition is not satisfied by the path setting when the revenue loss becomes larger than the revenue obtained by setting the path on the link. To do.
(4) Out of the procedure for detecting the path setting request including at least the transmission source node and the destination node and the quality class, and the condition of the bandwidth restriction corresponding to the quality class is satisfied among the routes from the transmission source node to the destination node. And a procedure for setting a path to a route with the lowest cost.

According to the present invention, the following effects are achieved.
(1) A path setting request in which the loss of revenue that is lost by setting a path on a link is greater than the revenue obtained by setting a path on the link. The link bandwidth constraint that minimizes the long-term profit loss caused by the set request loss is required.
(2) Since the bandwidth constraint is regularly updated based on the measurement result of the path usage bandwidth, the bandwidth constraint can always be maintained at the optimum value.
(3) Of the routes from the source node to the destination node, the path is set to the route with the minimum cost that satisfies the bandwidth constraint condition according to the quality class for each link. It is possible to set a path to an optimum route that can suppress a long-term profit loss caused by a set request loss.

  DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the best embodiment of the present invention will be described in detail with reference to the drawings. FIG. 1 is a diagram showing a configuration of a main part of an MPLS (Multi Protocol Label Switching) network to which the link bandwidth restriction control of the present invention is applied, and a plurality of nodes N and a network management unit for managing each node N Includes M. Each link L between nodes has a bandwidth constraint BC (BC0, BC1, BC2 ...) for each path quality class, and a newly set path has a bandwidth constraint according to the quality class. Imposed.

  FIG. 2 is a flowchart showing the procedure of the bandwidth constraint setting method in the link bandwidth constraint control according to the present invention, which is executed in the network management unit M. Here, a case where there are three quality classes will be described as an example.

  In this embodiment, as will be described in detail below, the average used bandwidth of the path in each link is measured for each quality class at each predetermined bandwidth constraint setting period, and the path call is determined based on the measured value. An average value of the amount (number of path setting requests per unit time × average holding time) is estimated for each quality class. Based on the estimated average path call volume by quality class, the future revenue loss caused by setting a new path rather than the revenue obtained by setting a new path for each link. This is characterized in that the bandwidth restriction of each link is set / updated so that the path setting request is lost when the value is large.

  In step S1, the average used bandwidth bi (i = 0, 1, 2) of the path belonging to each quality class i (i = 0, 1, 2) is measured at each link on the network. The average used bandwidth of the paths belonging to each quality class is obtained by measuring the used bandwidth of the paths belonging to each quality class at every predetermined measurement period of the path used bandwidth and averaging these measured values. The calculation of the measured value may be performed by each node N and notify the calculation result to the network management unit M, or the network management unit M obtains the calculation result of the used bandwidth of each path from each node N, You may make it obtain | require by calculating the average value.

In step S2, it is determined whether or not the control period is a predetermined bandwidth restriction. If the control period is the control period, the process proceeds to step S3. In step S3, an average path call amount ai (i = 0, 1, 2) for each quality class applied to each link is estimated. This average path call volume ai by quality class is calculated based on the Markov nature of the path setup and release process (the nature that the future behavior is determined only by the current value and is independent of the past behavior). For each band constraint method, it is estimated as a nonlinear solution of the following three simultaneous equations (1)-(3), (4)-(6), (7)-(9).
[For RDM]

[MAMの場合]

[For MAM]

[MARの場合]

[For MAR]

  In step S4, future revenue loss corresponding to each link state is calculated. Here, for the sake of simplicity, description will be made on the assumption that the bandwidth of the path is 1.0 in all quality classes, and the average hold time of the paths is the same in all quality classes. Further, the average holding time of the path is considered as a unit time.

  Assuming that the number of accommodated paths ni (i = 0, 1, 2) belonging to each quality class is a link state, assuming that there is no bandwidth limitation, there are paths in each of the quality classes 0, 1, 2 The probability of accommodating n0, n1, and n2, that is, the link state probability S (n0, n1, n2) is given by the following equations (10) and (11).

  If the future profit loss per unit time when starting from each link state is V (n0, n1, n2), the following formula is established for each link state by Markov decision theory.

  That is, it is assumed that the arrival of the path setting request and the release of the setting path follow a Markov process, and consider a sufficiently short time Δt. At this time, since the average hold time of all quality class paths is 1.0, the end rate per unit time of the set paths (per one) of all quality classes is 1.0, and each quality class The average path call volume (estimated value) is the quality class path setting request arrival rate per unit time.

  Here, starting from the state (n0, n1, n2), and assuming that the profit loss until kΔt time elapses is V (k) (n0, n1, n2), consider the case of [n0 + n1 + n2 <BW] State (n0, n1, n2) is during time Δt,

Transition to state (n0 + 1, n1, n2) with probability a0Δt,
Transition to state (n0, n1 + 1, n2) with probability a1Δt,
Transition to state (n0, n1, n2 + 1) with probability a2Δt,
Transition to state (n0-1, n1, n2) with probability n0Δt,
Transition to state (n0, n1-1, n2) with probability n1Δt,
Transition to state (n0, n1, n2-1) with probability n2Δt,
Stay in state (n0, n1, n2) with probability (1-a0Δt-a1Δt-a2Δt-n0Δt-n1Δt-n2Δt).
Therefore, the profit loss V (k) (n0, n1, n2) can be expressed by the following equation (12).

  When this is modified, the following equation (13) is obtained.

  Here, V (k) (n0, n1, n2) is the revenue loss from the state (n0, n1, n2) until kΔt time elapses, and V (k-1) (n0, n1, n2) is Since it is a profit loss from the state (n0, n1, n2) until (k-1) Δt time elapses, the difference between the two does not depend on the initial state (n0, n1, n2) if k is sufficiently large , Δt is the average revenue loss per hour. Therefore, if k is sufficiently large, the following equation (14) is established. However, RevLoss is the average revenue loss per unit time.

  In the above equation (13), k is sufficiently large, the left side is replaced with RevLoss * Δt, and k is sufficiently large, so V (k-1) (,,) is replaced with V (,,), and both sides are Δt By dividing by the following equation (15), the average profit loss RevLoss is obtained.

  On the other hand, considering the case of [n0 + n1 + n2 = BW], during the time Δt from the state (n0, n1, n2),

Revenue loss Rev0 occurs with probability a0Δt and stays in the state (n0, n1, n2)
Revenue loss Rev1 occurs with probability a1Δt and stays in the state (n0, n1, n2)
Revenue loss Rev2 occurs with probability a2Δt and stays in the state (n0, n1, n2)
Transition to state (n0-1, n1, n2) with probability n0Δt,
Transition to state (n0, n1-1, n2) with probability n1Δt,
Transition to state (n0, n1, n2-1) with probability n2Δt,
Stay in state (n0, n1, n2) with probability (1-a0Δt-a1Δt-a2Δt-n0Δt-n1Δt-n2Δt).

  Therefore, the following equation (16) is established. However, the revenue obtained from the quality class i setting pass is Revi.

  If this is modified in the same manner as described above, the following equation (17) is obtained.

  Here, ignoring the term whose value is zero, for example, by setting V (0, 0, 0) = 0 and solving the above simultaneous linear equations, the revenue loss V (n0, n1 in each link state , n2) is obtained.

Returning to the flowchart of FIG. 2, in step S5, the path setting request is lost when the future profit loss caused by setting the path is larger than the profit obtained by setting the path. A bandwidth constraint BC is required. The method for obtaining the bandwidth constraint in each bandwidth constraint model is as follows.
[For RDM]
BC0 = BW.

  The above formula represents that the future increase in profit loss (left side) due to setting a path exceeding the link bandwidth constraint is larger than the profit (right side) obtained by setting the path. In addition, the left side shows the future profit loss difference (expressed as the difference between two V (,,)) when the path is set beyond the link bandwidth constraint and the probability that each case will occur Expressed in weighted average form with S (,,).

  Here, the minimum value of BC1 ′ (<BW) that satisfies this inequality is defined as BC1. If BC1 ′ (<BW) satisfying the above inequality does not exist, BC1 = BW.

The minimum value of BC2 ′ (≦ BC1) that satisfies the above inequality (19) is defined as BC2. When there is no BC2 ′ (≦ BC1) that satisfies the above inequality, BC2 = BC1.
[For MAM]

  Let BC0 be the minimum value of BC0 ′ (<BW) that satisfies the above inequality (20). If BC0 ′ (<BW) that satisfies the above inequality does not exist, BC0 = BW.

  Let BC1 be the minimum value of BC1 ′ (<BW) that satisfies the above inequality. When there is no BC1 ′ (<BW) that satisfies the above inequality (21), BC1 = BW.

Let BC2 be the minimum value of BC2 ′ (<BW) that satisfies the above inequality (22). If there is no BC2 ′ (<BW) that satisfies the above inequality, then BC2 = BW.
[For MAR]

  The minimum BC0 ′ (≦ BW−RBW) satisfying the above inequality (23) is defined as BC0. When BC0 ′ (≦ BW−RBW) satisfying the above inequality does not exist, BC0 = BW−RBW.

  Let BC1 be the minimum value of BC1 ′ (≦ BW−RBW) that satisfies the above inequality (24). When there is no BC1 ′ (≦ BW−RBW) that satisfies the above inequality, BC1 = BW−RBW.

  The minimum BC2 ′ (≦ BW−RBW) satisfying the above inequality (25) is defined as BC2. When BC2 ′ (≦ BW−RBW) satisfying the above inequality does not exist, BC2 = BW−RBW.

  Next, a path setting method according to the second embodiment of the present invention will be described with reference to the flowchart of FIG. In the present embodiment, a path is set between nodes that have been requested to set a path by regarding the minimum cost path that satisfies the bandwidth constraint in each link as an optimum path.

  When a path setting request including information specifying the source node and destination node pair and the quality class is detected in step S20, the quality class of the path to be set for each link in the network is detected in step S21. Accordingly, the link cost is calculated based on one of the following formulas (26), (27), and (28). However, the cost of a link that does not satisfy the bandwidth constraint by setting a path is unconditionally regarded as infinite.

(The cost of a link whose status when setting a quality class 0 path is (n0, n1, n2))

（品質クラス1のパスを設定する時の状態が(n0, n1, n2)であるリンクのコスト）

(The cost of a link whose status when setting a quality class 1 path is (n0, n1, n2))

（品質クラス2のパスを設定する時の状態が(n0, n1, n2)であるリンクのコスト）

(The cost of a link whose status when setting a quality class 2 path is (n0, n1, n2))

  In step S22, the minimum cost route is determined as the optimum route based on the link cost. Note that when there is no route whose cost is not infinite, a path setting request is lost. In step S23, a path is set to the determined optimum route.

  According to the present embodiment, since a path can be set to an optimum route from the viewpoint of route cost, a long-term profit loss caused by a path setting request loss can be suppressed to be low in the entire network.

It is the block diagram which showed the structure of the MPLS network to which the link bandwidth restriction control of invention is applied. 5 is a flowchart showing a procedure of link bandwidth restriction control according to the present invention. It is the flowchart which showed the procedure of the route search method which concerns on 2nd Embodiment of this invention. FIG. 3 is a diagram schematically representing a bandwidth restriction method of RDM. It is the figure which expressed typically the bandwidth restriction method of MAM.

Explanation of symbols

N ... multiple nodes
M ... Network Management Department

Claims (3)

In a link bandwidth constraint control method for setting a quality difference for each path by providing a bandwidth constraint according to a path quality class on each link of the network,
A procedure for measuring the bandwidth used by each quality class for each link;
A procedure for calculating an average path call volume for each of the quality classes based on the measurement result of the path use bandwidth;
A procedure for obtaining, based on the average path call volume , revenue for each quality class in the future by newly setting a path for each link;
Based on the average path call volume , a procedure for obtaining a profit loss for each quality class that will be damaged in the future by newly setting a path for each link;
For a path setting request in which revenue loss is greater than the revenue, a procedure for setting bandwidth restrictions to be assigned to each quality class so that the bandwidth restriction condition is not satisfied by the path setting;
A bandwidth for link bandwidth constraint control, including a procedure for updating bandwidth constraints assigned to each quality class by repeating each procedure from calculation of path call volume to bandwidth constraint setting for each predetermined control period. Constraint setting method. In a path setting method for setting a path to an optimum route from a source node to a destination node on a network in which a quality difference is set for each path by providing a bandwidth restriction corresponding to the quality class of the path in the link,
A procedure for measuring the bandwidth used by each quality class for each link;
A procedure for calculating an average path call volume for each of the quality classes based on the measurement result of the path use bandwidth;
A procedure for obtaining, based on the average path call volume , revenue for each quality class in the future by newly setting a path for each link;
Based on the average path call volume , a procedure for obtaining a profit loss for each quality class that will be damaged in the future by newly setting a path for each link;
For a path setting request in which revenue loss is greater than the previous revenue, a procedure for setting bandwidth constraints to be assigned to each quality class so that the bandwidth constraint condition is not satisfied by the path setting;
A procedure for updating the bandwidth constraint assigned to each quality class by repeating each procedure from the calculation of the path call volume to the setting of the bandwidth constraint for each predetermined control period ;
A procedure for detecting a path setup request including at least a source node and a destination node, and a quality class;
Depending on the quality class of the path you are trying to set up, calculate the cost of each link,
A path setting method comprising: setting a path to a minimum cost route that satisfies a bandwidth constraint condition according to the quality class. As the cost of each link, the bandwidth constraint is satisfied by using the difference between the future revenue loss when a path is newly set and the future revenue loss when the path is not set, and by setting the path. 3. The path setting method according to claim 2 , wherein the link cost is regarded as infinite when the link cost is lost.

Images