JP4586183B2 - BAND MANAGEMENT METHOD, BAND MANAGEMENT DEVICE, AND PROGRAM FOR THE SAME IN COMMUNICATION NETWORK - Google Patents

Description

  The present invention relates to a bandwidth management technique in an IP (Internet Protocol) network, and in particular, uses a packet information sampled within a predetermined measurement period to identify a higher-level flow group that generates traffic occupying a link use bandwidth. The present invention relates to a bandwidth management method, a bandwidth management device, and a program therefor.

  As IP networks are widely used, there is an increasing demand for communication quality assurance on IP networks. On the other hand, the traffic characteristics of each flow are becoming more and more diverse, as represented by the abrupt changes in traffic patterns associated with the emergence of peer-to-peer (P2P) applications. ing.

  For example, when it is required to appropriately control a flow that occupies a large band for a long time such as P2P traffic and to ensure the quality of a flow with a small file size such as web sensitive to response time Is assumed. On the other hand, it is necessary to reduce the cost for communication facilities. Accordingly, it is important to appropriately control the use band of each flow so that the desired communication quality of each flow can be maintained by effectively using the given communication band. For that purpose, a method for identifying the long flow is required.

  As conventional techniques, there are methods for monitoring and controlling the state of all flows (for example, WFQ (Weighted Fair Queueing) in Non-Patent Document 1 and CSFQ (Core-Stateless-Fair-Queueing) in Non-Patent Document 2). However, with this method, it is necessary to manage the state of each flow at all nodes or network boundary nodes, and considering the situation where the number of flows will increase as the line speed increases in the future, There was a problem that it was not possible to deal with.

  On the other hand, as a means of avoiding such a problem, when the port number of the packet is viewed and it can be identified from the port number that the flow is related to the P2P type file sharing application, such a flow transfers a large file. Therefore, since it can be determined that there is a high possibility of occupying a band for a long time, a method for controlling only such a flow is also conceivable.

  However, since most P2P applications in recent years occupy most of the traffic that cannot be identified by the port number, this method cannot be used. In addition to P2P, DDoS (Distributed Denial of Service) is an attack in which a large number of computers distributed over multiple networks send packets to a specific server all at once, overflowing the communication path and stopping the function. In this way, abnormal traffic that continues to generate traffic at a high rate is increasing, and a method that does not depend on an application is required.

  On the other hand, packet sampling is attracting attention as a scalable measurement method as the line speed increases. For example, Non-Patent Document 3 is a long flow identification method using sampling. In this method, it is possible to specify the flow that sends out the number of packets exceeding a predetermined threshold only by sampling information, but how much bandwidth the specified flow group actually has on the line It was difficult to grasp even information such as whether it was occupied.

  However, such information regarding the bandwidth usage rate is very important in actually controlling the bandwidth. This is because it is impossible to determine how many upper flows should be controlled without knowing how many flows are used and how much bandwidth is to be controlled.

AK Parekh and RG Gallager, “A Generalized Processor Sharing Approach to Flow Control in Integrated Services Networks: The Single-Node Case,” IEEE / ACM Trans. On Networking, vol. 1, no. 3, (June 1993) pp.344 -357. I. Stoica, S. Shenker, and H. Zhang, “Core-stateless fair queuing: achieving approximately fair bandwidth allocations in high-speed networks,” proceedings of ACM SIGCOMM98, pp. 118-130, 1998. T. Mori, M. Uchida, R. Kawahara, J. Pan, and S. Goto, “Identifying elephant flows through periodically sampled packets,” ACM SIGCOMM Internet Measurement Conference, 2004.

  An object of the present invention is to provide a bandwidth management method capable of maintaining the communication quality of each flow by appropriately controlling a long flow group without performing state management for each flow at each node in view of the above-described problems, To provide a bandwidth management device and a program therefor.

  In the present invention, a packet sampling technique is used to reduce the number of management flows. This technique will be described with reference to FIG.

  Consider the case where packets from multiple flows arrive as shown in FIG. As an example of the flow definition, here, a packet group having the same (source IP address, destination IP address, source port number, destination port number, protocol number) is defined as a flow.

  In FIG. 1, each packet is shown in a pattern (pattern) divided by flow. Here, if one packet is sampled every N (sampling rate f = 1 / N) and only the flow to which the sampled packet belongs is managed, the number of management flows is reduced to two in the case of FIG. Yes (5 if not sampled).

  In this way, the number of management flows can be reduced by sampling. As can be seen from FIG. 1, a flow having a larger number of packets is more easily sampled. Therefore, what is desired to be controlled now is a flow that occupies a band for a long time, that is, a flow with a large number of packets, and therefore it is sufficient to target only a flow in which packets are sampled.

a) In the first bandwidth management method of the present invention, a packet is sampled in a node of a communication network, information on a flow to which the sampled packet belongs is managed, and a bandwidth used for each flow in an output link from the node is determined. In a bandwidth management method using a managing computer, a step of sampling packets at a sampling rate f within a predetermined measurement period T, and a group of upper flows related to a link bandwidth using the packet information sampled in the step It has the step which specifies.

In the first bandwidth management method of the present invention, the step of specifying the upper flow group further includes measuring the number of packets Yj sampled from the flow j and sorting the flow j in descending order with respect to Yj. The total number of packets is n (that is, n = ΣYj), and random sampling extraction (Random Sampling With Replacement) is performed n times each time a random number is extracted. And a step of estimating the number of packets sampled from the flow j by the random restoration extraction as Zj and the original number of packets included in the flow j by Wj = 1 / f × min {Yj, Zj} Estimating the average packet size Lj of flow j by Lj = {total number of bytes of packets sampled from flow j} / Yj; A step of measuring the use bandwidth A [bps], and obtaining the minimum m satisfying Σj = 1 to m (Wj × Lj)> x / 100 × AT (x is a predetermined target value) It is characterized by comprising the steps of specifying up to m as upper flow groups occupying x% of the link bandwidth.

Here, a method for estimating the original number of packets of the flow j will be described. A method of setting Wj = 1 / f × Yj as in a second method described later is also conceivable, but here, the method described above is used. As described above, it is possible to improve the estimation accuracy by performing random restoration extraction from the sampled packets to create a new number of packets Zj, and estimating Wj using that information as well. This is shown using actual data in FIG.

b ) The second bandwidth management method of the present invention replaces the step of estimating the original number of packets included in the flow j with Wj = 1 / f × min {Yj, Zj} in the first bandwidth management method. And the step of estimating the original number of packets included in the flow j by Wj = 1 / f × Yj.

c ) The third bandwidth management method of the present invention is the first bandwidth management method in which random restoration extraction is performed n times using the sampled packets as a population, and from the flow j by random restoration extraction. Instead of estimating the number of sampled packets as Zj and the original number of packets contained in flow j by Wj = 1 / f × min {Yj, Zj}, the sampled packets are randomly restored as a population. A step of repeatedly performing d sets of processes for performing extraction n times, and the number of packets sampled from the flow j as Zj, and the number of original packets included in the flow j as Wj = 1 / f × min {Yj, Zj / d} or Vj ^ {(i)} = min (Yj, Zj ^ {(i)}), (where i = 1, 2,... D, where Zj ^ {(i )} Is sampled on the i th z The number of packets of the flow j), arranging them in descending order, extracting a predetermined k-th (1 ≦ k ≦ d), and estimating by Wj = 1 / f × Vj ^ {(k)} It is characterized by having.

d ) The fourth bandwidth management method of the present invention measures the current link usage bandwidth A [bps], sets a predetermined target link usage bandwidth as C ^*, and if A> C ^* , claims 1 to 4 A flow management method according to any one of the above, the upper flow group related to the used bandwidth is specified, and the minimum m satisfying Σj = 1 to m (Wj × Lj)> (AC ^* ) T is obtained; From m to m, and two types of queues waiting for output to the link are prepared at the node, and steps 1 to m are accommodated in the low priority queue and the rest are accommodated in the high priority queue. Features.

e ) The fifth bandwidth management method of the present invention is such that, in the fourth bandwidth management method, instead of accommodating the flows 1 to m in the low priority queue, the upper flow group is diverted to another route. A bandwidth management method characterized by the above.

f ) In the sixth bandwidth management method of the present invention, the upper flow group is controlled by the fourth or fifth bandwidth management method, and the actual arrival byte number of the flow group to be controlled and the first and first flow groups are controlled. The number of arrival bytes Σj = 1 to m (Wj × Lj) estimated by the bandwidth management method 2 is compared, and the control target flow in the next cycle is determined as the fourth or 5 is determined and controlled using the bandwidth management method.

This is done by performing both the first estimation method and the second estimation method at the same time, and by measuring the number of true arrival bytes from the upper flow group actually controlled in the previous cycle ( When controlling, collect all packet information as well as sampling information for these flow groups), determine which estimation method is appropriate at the present time, and specify the control target flow group in the next cycle. An estimation method is used.

g) seventh band management apparatus of the present invention is to sample the packet at a node of a communication network, and manages the information flow that belongs the sampled packets, the bandwidth used for each flow at the output link from the node In a bandwidth management device using a managing computer, a means for sampling packets at a sampling rate f within a predetermined measurement period T, and a higher-level flow group related to a link bandwidth using packet information sampled by the means It has the means to specify.

In the seventh bandwidth management apparatus of the present invention, the means for specifying the upper flow group further includes a means for measuring the number of packets Yj sampled from the flow j and sorting the flow j in descending order with respect to Yj. The total number of packets is n, means for performing random restoration extraction n times with the sampled packets as a population, and the number of packets sampled from the flow j by the random restoration extraction is Zj, and included in the flow j Means for estimating the number of original packets by Wj = 1 / f × min {Yj, Zj} and the average packet size Lj of flow j by Lj = {total number of bytes of packets sampled from flow j} / Yj A means for estimating, a means for measuring the link bandwidth A [bps], and a maximum satisfying Σj = 1 to m (Wj × Lj)> x / 100 × AT (where x is a predetermined target value). Of the calculated m, characterized in that it comprises means for specifying a higher flow group accounts for x% of the link bandwidth used from flow 1 to m.

h ) The eighth bandwidth management apparatus of the present invention is the seventh bandwidth management apparatus, instead of means for estimating the original number of packets included in the flow j by Wj = 1 / f × min {Yj, Zj}. , And having means for estimating the number of original packets included in the flow j by Wj = 1 / f × Yj.

i ) The ninth bandwidth management apparatus of the present invention is a seventh bandwidth management apparatus comprising means for performing random restoration extraction n times using a sampled packet as a population, and a flow j by random restoration extraction. Randomly restore sampled packets as population instead of means to estimate the number of packets sampled as Zj and the original number of packets contained in flow j by Wj = 1 / f × min {Yj, Zj} Means for repeatedly performing d sets of processes for performing extraction n times, and the number of packets sampled from the flow j as Zj, the original number of packets included in the flow j as Wj = 1 / f × min {Yj, Zj / d} or Vj ^ {(i)} = min (Yj, Zj ^ {(i)}), (where i = 1, 2,... D, where Zj ^ {(i )} Is the flow j sampled on the i th jet Characterized in that they are arranged in descending order, a predetermined k-th is extracted (1 ≦ k ≦ d), and estimated by Wj = 1 / f × Vj ^ {(k)}. And

j ) The tenth bandwidth management apparatus of the present invention is the ninth bandwidth management apparatus, in which flows 1 to m are controlled, two types of output waiting queues for the link are prepared in the node, It is characterized by having means for accommodating upper flow groups 1 to m in a low priority queue or detouring to another route and accommodating the rest in a high priority queue.

k ) The bandwidth management program of the present invention is a bandwidth management program for causing a computer to function as each means in the bandwidth management device according to any one of Items 7 to 10 .

  According to the present invention, it is possible to appropriately detect a flow that generates bandwidth occupying traffic without performing state management for each flow at each node, and to improve the response time of a small-sized flow such as web. It is possible to provide a bandwidth management method and apparatus and a bandwidth management program that enable effective use of the bandwidth.

  Embodiments of the present invention will be described below with reference to the drawings.

FIG. 2 is a configuration diagram showing an example of a basic configuration to which the present invention is applied.
As shown in FIG. 2, the bandwidth management apparatus 20 of the present invention is constructed in such a manner that it is inserted between the nodes in the IP network (between the nodes 10 and 30 in FIG. 2). Alternatively, it may be externally attached to a node, and the packet may be mirrored to that node.

FIG. 3 is a block diagram showing a configuration example of a bandwidth management device using the second method of the present invention.
As shown in the figure, the bandwidth management apparatus 20 according to the present invention includes a packet size reading unit 21, a link usage bandwidth calculation unit 22, a usage bandwidth management unit 23, a sampling unit 24, a packet header analysis unit 25, and a flow management unit 26. The flow packet number estimation unit 27 is included. The sampling unit 24 includes a counter 241, and the flow management unit 26 includes a flow management table 261.

  The bandwidth management device 20 has a normal computer configuration and controls hardware such as a CPU (Central Processing Unit), a storage device, an external storage device such as a hard disk, an input / output unit such as a keyboard and a mouse, and the like. At the same time, the CPU implements the functions of the packet size reading unit 21, the link usage band calculation unit 22, the usage band management unit 23, the sampling unit 24, the packet header analysis unit 25, the flow management unit 26, and the flow packet number estimation unit 27. And a program for making it happen.

FIG. 4 is a flowchart of processing performed by the bandwidth management apparatus according to the present invention.
The processing of the bandwidth management apparatus according to the present invention will be described below with reference to the block diagram of FIG. 3 and the flowchart of FIG.
As shown in the block diagram of FIG. 3, when a packet arrives from the preceding node, the packet size reading unit 21 reads the packet size of the packet, and the value is notified to the link use band calculation unit 22 (step S101), the link use band calculation unit 22 calculates the current link use band A by dividing the sum of the sizes of the packets passed during the predetermined time T by T, and notifies this to the use band management unit 23 (S101). Step S102).

  On the other hand, the arrived packet is transferred to the sampling unit 24, and the sampling unit 24 counts up the value C of the counter 241 and checks whether the value C of the counter 241 becomes equal to a predetermined sampling period N ( Step S103).

  When the value C of the counter 241 becomes equal to a predetermined sampling period N (C = N), the packet is transferred to the packet header analysis unit 25, and the counter 241 is reset (C = 0). The data is transferred to the subsequent node (step S104). If the value of the counter 241 is less than the predetermined sampling period N (C <N), the packet is simply transferred only to the subsequent node (step S105).

  As an alternative method, instead of comparing the value C of the counter 241 with the sampling period N, a random number is generated in the range of 0 to 1 for each packet arrival. The packet may be transferred to the subsequent node after being transferred to the packet header analysis unit, otherwise the packet may be transferred only to the subsequent node.

  The packet header analysis unit 25 reads (source IP address, destination IP address, source port number, destination port number, protocol number, packet size) and notifies the information to the flow management unit 26 (step S106). ).

  The flow management unit 26 prepares in advance a flow management table 261 for managing the state for each flow. As shown in FIG. 5, the flow management table 261 stores the arrival packet number Yj and the arrival byte number Bj from the flow j for each flow (step S107).

  When the packet is a packet from a new flow, the flow is entered in the flow management table 261, Yj ← 1 and Bj ← Pkt. Here, Pkt is the packet size. If the packet is from a flow j that has already been entered, update to Yj ← Yj + 1, Bj ← Bj + Pkt.

  The above procedure is performed until a predetermined time period T elapses after the measurement is started. At the time T elapses, the flow order is rearranged in descending order with respect to Yj, and then the values of Yj and Bj are changed to flow packets. The number estimation unit 27 is notified. After the notification, these values are reset (step S108).

  The flow packet number estimation unit 27 calculates the total number n of sampled packets by n = ΣYj, generates a uniform random number from 0 to 1, and the value X is Σk = 1 to j−1 (Yk / n ) ≦ X <Σk = 1 to j (Yk / n), Zj ← Zj + 1. This is repeated n times and Zj is updated (step S109).

  Thereafter, the original number of packets of the flow j is estimated by Wj = 1 / f × min {Yj, Zj}. Here, f is a sampling rate, and f = 1 / N. Further, the packet size Lj of the flow j is estimated by Lj = Bj / Yj. This result is notified to the used bandwidth management unit 23 (step S110).

  The used bandwidth management unit 23 uses the current used bandwidth A of the entire link notified from the link used bandwidth calculating unit 22, the number of packets Wj of the flow j notified from the flow packet number estimating unit 27, and the packet size Lj. , Σj = 1 to m (Wj × Lj)> x / 100 × AT is determined as the minimum m (x is a predetermined target value), and flows 1 to m occupy the used bandwidth x% It identifies as a flow group (step S111).

  According to the first embodiment, it is possible to specify the upper flow group only from the sampling information.

  In the first embodiment, instead of estimating the original number of packets included in the flow j by Wj = 1 / f × min {Yj, Zj}, Wj = 1 / f × Yj.

  According to the second embodiment, it is not necessary to perform non-restoration extraction again as in the first embodiment, and the processing load can be reduced.

  In the flow packet number estimation unit of the first embodiment, Zj was calculated by n trials (this is assumed to be one set), but this was repeated d sets, and the original number of packets of flow j was determined. Estimation is performed using Wj = 1 / f × min {Yj, Zj / d}. Alternatively, Vj ^ {(i)} = min (Yj, Zj ^ {(i)}), (where i = 1, 2,... D, where Zj ^ {(j)} is the i th The number of packets of flow j sampled in the zette), arranging them in descending order, extracting the predetermined k-th (1 ≦ k ≦ d), and Wj = 1 / f × Vj ^ {(k)} presume.

  According to the third embodiment, it is possible to improve estimation accuracy by repeatedly performing non-restoration extraction for d sets.

In the first embodiment, the operator is notified of the upper flow group. Here, however, two types of output waiting queues for the link are prepared at the node, and a predetermined link use bandwidth target value C is set. ^* Is used, if the current bandwidth A is greater than C ^{*, the} minimum m satisfying Σj = 1 to m (Wj × Lj)> (A−C ^* ) T is obtained, and the flow from 1 to m is reduced. It is accommodated in the priority queue and the rest is accommodated in the high priority queue, or flows 1 to m are routed through another route.

  According to the fourth embodiment, it is possible to automatically control the upper flow group (accommodating in the low priority queue or via another route) without using an operator.

  The upper flow group is controlled by the method of the fourth embodiment, and at the same time, the true arrival byte number of the flow group to be controlled is measured. At the same time, the arrival byte number of the flow group is calculated by the method of the first and second embodiments. The control target flow in the next cycle is estimated using the estimation method in which the estimation value is estimated by = 1 to m (Wj × Lj) and the estimation value is close to the actual measurement value by comparing the true value with the estimation values 1 and 2. To determine and control.

  According to the fifth embodiment, it is determined which of the estimation methods of the first and second embodiments is appropriate at the present time, and an appropriate estimation method is used for specifying the control target flow group in the next cycle. It can be used.

  In the first embodiment, the bandwidth management device 20 is realized by being inserted between the nodes. Instead, as shown in FIG. 6, sampling packet information is collected from each node constituting the network. The same management may be performed for the entire network.

  According to the sixth embodiment, the bandwidth management technology according to the present invention can be extended to the monitoring of the entire network.

In the first embodiment, the number of sample packets of each flow is counted every predetermined measurement period T, but each time M packets arrive at the node (M is a predetermined number of packets). The number of sample packets in the flow may be counted.
In the first embodiment, the packet size reading unit reads the size of all packets and calculates the use band A. On the other hand, using the Yi, Bi (number of bytes of flow i) in the flow management unit, ΣB i × AT (a value obtained by multiplying the use band A by the measurement period T) may be estimated by M / ΣYi. Alternatively, the AT may be estimated by ΣB i × 1 / f.
In the first embodiment, the original number of packets is estimated by Wj = 1 / f × min {Yj, Zj}, whereas it is estimated by Wj = M / ΣYi × min {Yj, Zj}. Good.
In this way, all packet processing in the packet size reading unit can be made unnecessary.

  Next, the effect in this invention is demonstrated using the result confirmed using real data. In this example, data obtained by packet capture on an internet backbone line is used.

FIG. 7 shows the result of packet sampling performed at a sampling rate of f = 1/1000 from the original number of packets of 10 ⁷ and specifying the upper flow group of the used bandwidth x% by the method of the first embodiment. . The number of experiments is 100, and the average of the estimated values and the confidence interval at that time are shown (both units are [× 10 ³ bytes]).

  In FIG. 7, the first column indicates the percentage of the link bandwidth, the second column is the number of upper flows occupying x% (the number of flows specified in the present invention), and the third column is the flow group occupying x%. True value of bandwidth usage, 4th column is the average of estimated bandwidth usage of the flow group occupying x% (estimation according to the present invention), 5th column is estimation of bandwidth usage of the flow group occupying x% The lower 2.5% value is the lower 2.5% value, and the sixth column is the upper 2.5% value of the estimated band usage of the flow group occupying x%.

  As can be seen from FIG. 7, the above estimation result falls between 884, 480-904, and 444 with 95% reliability, and true values 894 and 462 are included in the section. Therefore, it is estimated OK. This confirms that the estimation method in this example is appropriate.

  Note that a program for realizing the present invention (that is, a packet size reading unit 21, a link use band calculation unit 22, a use band management unit 23, a sampling unit 24, a packet header analysis unit 25, a flow management unit 26, and the number of flow packets) A program for realizing various functions of the estimating unit 27) can be widely distributed in the market via CD-ROM, DVD, FD, and the Internet.

It is a figure explaining the operation | movement of the packet sampling used by this invention. It is a block diagram which shows an example of the basic composition to which this invention is applied. It is a block diagram of a bandwidth management device in the present invention. It is a flowchart of the process which the bandwidth management apparatus in this invention performs. It is a structural example of the flow management table of the flow management part in this invention. It is a figure which shows another embodiment of the band management apparatus in this invention. It is a figure for demonstrating the effect of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10: Node 20: Band management apparatus 21: Packet size reading part 22: Link use band calculation part 23: Use band management part 24: Sampling part 241: Counter 25: Packet header analysis part 26: Flow management part 261: Flow management table 27: Flow packet number estimation unit 30: Node

Claims (11)

In a bandwidth management method using a computer that samples a packet in a node of a communication network, manages information of a flow to which the sampled packet belongs, and manages a bandwidth used for each flow in an output link from the node.
Possess a step of sampling the packets at a sampling rate f in a predetermined within the measurement period T, by using the packet information sampled in said step, a step of identifying the upper flow group Links used band,
The step of identifying the upper flow group includes:
Measure the number of packets sampled Yj from flow j, sort flow j in descending order with respect to Yj, and set the total number of packets sampled to n, and use random sampling for sampling And a step of estimating the number of packets sampled from the flow j by the random restoration extraction as Zj and the original number of packets included in the flow j by Wj = 1 / f × min {Yj, Zj} , Estimating the average packet size Lj of the flow j by Lj = {total number of bytes of packets sampled from the flow j} / Yj, measuring the link bandwidth A [bps], and Σj = 1˜ m (Wj × Lj)> x / 100 × AT (where x is a predetermined target value) is determined as the minimum m, and flows 1 to m are x of the link usage bandwidth. Bandwidth management method characterized by comprising the step of identifying as a higher flow group occupying. In the bandwidth management method as claimed in claim 1,
Instead of estimating the number of original packets included in the flow j by Wj = 1 / f × min {Yj, Zj}, the number of original packets included in the flow j is estimated by Wj = 1 / f × Yj. A bandwidth management method comprising the steps of: The bandwidth management method according to claim 1 ,
A step of performing random restoration extraction n times using the sampled packets as a population, and the number of packets sampled from the flow j by random restoration extraction as Zj, and the number of original packets included in the flow j as Wj = In place of the step of estimating by 1 / f × min {Yj, Zj}, a step of repeatedly performing a process of performing random restoration extraction n times with a sampled packet as a population, d sets, and from flow j Zj is the number of sampled packets, and the number of original packets included in flow j is estimated by Wj = 1 / f × min {Yj, Zj / d}, or Vj ^ {(i)} = min (Yj , Zj ^ {(i)}), (where i = 1, 2,... D, where Zj ^ {(i)} is the number of packets of flow j sampled in the i th jet) These are arranged in descending order and determined in advance. k-th withdrawal (1 ≦ k ≦ d), the bandwidth management method characterized by comprising the step of estimating by Wj = 1 / f × Vj ^ {(k)}. Measuring the current link usage band A [bps], the target link using a predetermined band and C ^*, the use band in a band management method according to any one of A> C ^* If claims 1 to 3 Identify the upper flow group, calculate the minimum m satisfying Σj = 1 to m (Wj × Lj)> (A−C ^* ) T, and control flow 1 to m as the control target. A bandwidth management method comprising the steps of: preparing two types of output waiting queues, storing upper flow groups of flows 1 to m in a low priority queue, and storing the rest in a high priority queue. The bandwidth management method according to claim 4 , wherein
A bandwidth management method characterized in that instead of accommodating flows 1 to m in a low priority queue, the upper flow group is detoured to another route. Controls the upper flow group in the band management method according to claim 4 or 5, which is estimated by the bandwidth management method as claimed actual number arrival of bytes of said flow group whose control target to claim 1 and 2 6. The bandwidth management method according to claim 4 or 5 , wherein the number of arrival bytes Σj = 1 to m (Wj × Lj) is compared, and the control target flow in the next cycle is estimated using an estimation method whose estimated value is close to the actual measured value. A bandwidth management method characterized by determining and controlling the bandwidth. In a bandwidth management apparatus using a computer that samples a packet in a node of a communication network, manages information of a flow to which the sampled packet belongs, and manages a bandwidth used for each flow in an output link from the node.
Possess a means for sampling the packets at a sampling rate f in a predetermined within the measurement period T, by using the packet information sampled in said means, a means for specifying a higher flow group Links used band,
The means for specifying the upper flow group is:
Measures the number of packets Yj sampled from the flow j, sorts the flow j in descending order with respect to Yj, and sets the total number of packets sampled to n, and uses the sampled packets as a population to perform random restoration extraction n times Means for carrying out, and means for estimating the number of packets sampled from the flow j by the random restoration extraction as Zj and the original number of packets included in the flow j by Wj = 1 / f × min {Yj, Zj}; , Means for estimating the average packet size Lj of the flow j by Lj = {total number of bytes of packets sampled from the flow j} / Yj, means for measuring the link bandwidth A [bps], and Σj = 1˜ m (Wj × Lj)> x / 100 × AT (where x is a predetermined target value) is determined as the minimum m, and flows 1 to m are upper flows that occupy x% of the link bandwidth. Band management apparatus which comprises a means for identifying as. The bandwidth management device according to claim 7 ,
Instead of means for estimating the number of original packets included in flow j by Wj = 1 / f × min {Yj, Zj}, the number of original packets included in flow j is estimated by Wj = 1 / f × Yj. A bandwidth management apparatus comprising means. The bandwidth management device according to claim 7 ,
A means for performing random restoration extraction n times with a sampled packet as a population, and the number of packets sampled from flow j by random restoration extraction as Zj, and the original number of packets included in flow j as Wj = In place of means for estimating by 1 / f × min {Yj, Zj}, means for repeatedly performing d sets of processes for performing random restoration extraction n times using a sampled packet as a population, and from flow j The number of sampled packets is set as Zj, and the number of original packets included in the flow j is estimated by Wj = 1 / f × min {Yj, Zj / d}, or Vj ^ {(i)} = min (Yj , Zj ^ {(i)}), (where i = 1, 2,... D, where Zj ^ {(i)} is the number of packets of flow j sampled in the i th jet) These are arranged in descending order, and the predetermined kth is extracted. Out (1 ≦ k ≦ d), the bandwidth management apparatus characterized by comprising means for estimating the ^ Wj = 1 / f × Vj {(k)}. 10. The bandwidth management apparatus according to claim 9 , wherein the flows 1 to m are controlled, two types of output queues for the link are prepared at the node, and the upper flow groups from the flows 1 to m are reduced. A bandwidth management apparatus comprising means for accommodating in a priority queue or detouring to another route and accommodating the rest in a high priority queue. A bandwidth management program for causing a computer to function as each means in the bandwidth management device according to any one of claims 7 to 10 .

Images