KR101643552B1 - Method and apparatus for estimating available bandwidth and capacity of network paths in communication systems - Google Patents

Abstract

A method for estimating available bandwidth and capacity of network paths in a communication system includes a parameter selection process for determining a size of an optimal packet chirp through repeated probing based on binary search between a minimum value and a maximum value of a packet chirp size, , The transmission speed and reception speeds of the determined plurality of packet colliders are compared, and if the reception speed is high, the probing speed is increased. If the reception speed is low, the probing speed is decreased and the point is stored. And estimating the bit rate capacity by determining a point at which the transmission speed is close to the available bandwidth as an anchor point and estimating an inclination between the anchor point and the remaining points after estimating the available bandwidth.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention [0001] The present invention relates to an apparatus and a method for estimating available bandwidth and capacity of network paths in a communication system,

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus and method for estimating a network path of a communication system, and more particularly, to an apparatus and method for estimating the available bandwidth and capacity of a network link in a network path.

In a communication system, the bandwidth of a network path is an important factor in the performance of many applications. Particularly, in a general communication environment in which many types of traffic share limited resources, a bottleneck link of a path is not directly related to application performance but also to QoS and flow control of traffic affect. A bottleneck link is a very complicated process in which the usable bandwidth among the links constituting the network path is the smallest link and the usable bandwidth and physical capacity of the link are measured without directly accessing the network resources.

Techniques for measuring bandwidth and capacity of a bottleneck link estimate a bandwidth by analyzing the transmission delay of packets received by sending a probe packet through a network path to be measured. Among these techniques, a technique based on a single-hop model assumes that a network path is composed of one link. Another technique is to measure a multi-hop path consisting of multiple links separately for each link.

Recently, a technique for estimating a bandwidth by utilizing a linear relationship between a transmission rate r _I for injecting a probe packet into a network and a reception rate r _O for arriving at a receiving end of the packet has been introduced. The technique estimates the available bandwidth A _t and the capacity C _t of the bottleneck link using a probing response curve, which is a functional relationship between the transmission rate r _I of the probe packet and the metric r _I / r _O.

In the prior art, a plurality of probe packet trains composed of a plurality of consecutive packets are transmitted at a specific transmission rate r _I , and then a packet reception rate r _O of the received packet traces is monitored to calculate r _I And r _I / r _O. This probing process should be repeated with varying transmission rates. Typically, the available bandwidth of a bottleneck link can range from zero to the capacity of that link (i.e., A _t ∈ [0, C _t ]). Therefore, the transmission rate r _I should start with a predetermined small value and proceed with increasing the r _I value every time the probing is repeated, and this process may also result in a second linear segment appearing in the response curve To increase the transmission speed.

The related art based on this probe response curve need only be accurately and rapidly determined response curve to the bandwidth estimation of the bottle neck link, but the response curve is a rise and fall in accordance with the change of the network and traffic conditions, and A _t, and Since there is no information on _Ct , the prior art estimation method is difficult to accurately and efficiently measure.

The present invention relates to a method and system for providing an available bandwidth of a bottleneck link in a network path that plays an important role in performance and quality of service (QoS) management of an application in a communication network such as WiMAX, Long-Term Evolution (LTE) and methods and apparatus for estimating available bandwidth and physical capacity.

The present invention relates to a multi-hop network environment using effective repetitive probing instead of full bandwidth inspection of the available bandwidth and capacity of a network path of a network path that plays an important role in application performance and QoS management in communication systems such as WiMAX and LTE To accurately estimate even under arbitrary traffic conditions.

According to the above-mentioned object, the present invention first determines a start value r _I of a transmission rate matching a path characteristic through simple probing using one packet reference. Next, the size N of the probe packet, which is variable according to the network conditions and the traffic conditions of the path, is determined through iterative probing based on the binary search using the probing start transmission rate obtained in the step. As described above, an automated algorithm that accurately estimates the available bandwidth and capacity of a bottleneck link without performing a full check by using the probe start transmission rate and packet size N, which match the path characteristics automatically selected, .

A method for estimating the available bandwidth and capacity of network paths in a communication system according to an embodiment of the present invention is a method for estimating the available bandwidth and capacity of network paths, Comparing the transmission speeds and reception speeds of the determined plurality of packet colliders to increase the probing speed when the reception speed is high and decreasing the probing speed and storing the points when the reception speed is low Estimating a range of the available bandwidth while repeating the estimation of the available bandwidth, determining a point at which the transmission speed is close to the available bandwidth as an anchor point, calculating an inclination between the anchor point and the remaining points, And estimating the capacity.

In addition, an apparatus for estimating available bandwidth and capacity of network paths in a communication system according to an embodiment of the present invention includes an apparatus for estimating available bandwidths and capacities of network paths, And a controller for comparing the transmission speeds and reception speeds of the determined plurality of packet colliders to increase the probing speed if the reception speed is high and decrease the probing speed if the reception speed is low, And determining, as an anchor point, a point at which the transmission rate is close to the available bandwidth after the estimation of the usable band-width range, and determining an anchor point between the anchor point and the remaining points And a capacity estimating unit for estimating a bucket neck capacity by obtaining a slope And that is characterized.

A method for estimating the available bandwidth and capacity of a network path in a communication system according to an embodiment of the present invention includes transmitting a packet packet consisting of a plurality of probe packets back-to-back , It is possible to easily determine an appropriate starting transmission rate r _I of the packet chirp to be used in probing by calculating the average reception rate of the packets in the packet packet and determining the value as the probing start transmission rate. In addition, the N selection block determines the optimal packet chirp size through the iterative probing based on the binary search between the minimum value (N min) and the maximum value (N max) of the packet chirp size, It is possible to accurately measure the reception speed of the probe packet chirp with the size N being transmitted at the determined transmission rate r _I. By using the probe start transmission rate and packet collision size N that match the path characteristics automatically selected, an automated algorithm that accurately estimates the available bandwidth A _t and the capacity C _t of the bottleneck link, Can be provided.

1 shows an example of an ideal probe response curve
2 is a diagram illustrating a process for estimating the available bandwidth and capacity of a bottleneck link according to an embodiment of the present invention.
FIG. 3 is a flow chart illustrating a procedure for determining a starting transmission rate and a size of a probe packet probe, which are probing parameters, in accordance with an embodiment of the present invention in FIG.
FIG. 4 is a flow chart illustrating the available bandwidth estimation procedure of a bottleneck link in accordance with an embodiment of the present invention in FIG.
FIG. 5 is a flow chart illustrating the available bandwidth estimation procedure of a bottleneck link in accordance with another embodiment of the present invention in FIG.
FIG. 6 is a flow chart illustrating the procedure for estimating the capacity of a bottleneck link in accordance with an embodiment of the present invention in FIG.
7 is a diagram illustrating an example of an architecture for performing an available bandwidth and capacity estimation of a bottleneck link according to an embodiment of the present invention

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. It should be noted that the same configurations of the drawings denote the same reference numerals as much as possible.

In the following description, specific details of each parameter (Nmin, Nmax, σ, k, η, etc.) are shown, which is provided only for better understanding of the present invention. It will be apparent to those skilled in the art that the present invention may be embodied in other forms without departing from the spirit and scope of the invention. In the following description of the present invention, detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear.

The present invention relates to a method and system for providing an available bandwidth of a bottleneck link in a network path that plays an important role in performance and quality of service (QoS) management of an application in a communication network such as WiMAX, Long-Term Evolution (LTE) and methods and apparatus for estimating available bandwidth and physical capacity.

To this end, the present invention first determines the start value r _I of the transmission rate according to the path characteristics through simple probing using one packet chirp. Next, the size N of the probe packet, which is variable according to the network conditions and the traffic conditions of the path, is determined through iterative probing based on the binary search using the probing start transmission rate obtained in the step. And, by using the probe start transmission rate r _I and the packet collision size N that match the path characteristics automatically selected as described above, an automated banding and capacity estimation of the bottleneck link can be accurately performed Algorithm.

Fig. 1 is a diagram showing, by way of example, the probe response curve F in the ideal case.

Referring to FIG. 1, the response curve F is composed of several piece-wise linear segments, and in the case of a multi-hop path, at least two linear segments are included. The first linear segment of the response curve F breaks off at the point where the probe transmission rate r _I equals the available bandwidth of the bottleneck link, A _t , to start the second segment. Point B at which the second segment is broken off is determined by the routing matrix of all traffic passing through the path, where probe transmission rate r _I is greater than the available bandwidth A _t of the bottleneck link.

If such a response curve can be obtained, it is not difficult to estimate the band-wise characteristic of the bottleneck link. That is, the available bandwidth A _t is the probe transmission rate at which the first breakpoint of the response curve F is generated. Then, the capacity C _t estimation of the bottleneck link is obtained by obtaining the slope of the second linear segment and taking the reciprocal thereof.

In this case, it is important to determine how to obtain the response curve F as shown in FIG. 1 through probing in a real network environment where the traffic conditions of the network and the path are unknown. That is, in order to obtain the response curve F as shown in FIG. 1, the following three conditions must be considered.

First, it is difficult to determine an appropriate starting transmission rate r _I of the packet chirp to be used for probing, because the available bandwidth of a bottleneck link, A _t , can be very wide depending on the degree of link utilization of the network traffic. If the start transmission rate r _{I is set} to a too small value, the probing time may be long, and unnecessarily large number of probing operations may be repeated, thereby increasing the amount of probe traffic injected into the network. On the other hand, using too large a starting transmission rate (for example, a value larger than A _t ) may result in failure to find the first breakpoint in the response curve and thus the bandwidth estimation error may become large. In the embodiment of the present invention, after transmitting one packet of the probe packet composed of a plurality of probe packets back-to-back, the average reception rate of the packets in the packet packet is calculated, Speed.

Second, in order to obtain an accurate response curve, the reception speed of the probe packet chirp with the size N to be transmitted at a different transmission rate r _I must be accurately measured. However, the measured value r _O of the received speed greatly depends on the traffic conditions of the path and the network conditions. Ideally, if the size of the probe packet is infinitely large or the size N of the probe packet train is infinite, a stable measurement value r _O for the transmission rate r _I of each packet chirp will be obtained. However, in a real network, the packet size is limited by the maximum transfer unit (MTU) of the network element, and the size of the packet chirp can not be increased without limitation because the buffer size of the router is finite . Therefore, it is not easy to determine an appropriate N value that is adaptive to the traffic environment of the network path and that can obtain a stable reception rate measurement value regardless of the transmission rate. For example, if the packet size N is too small to fit the network environment of the path too much, errors in the measured value r _O are increased and a reliable probe response curve can not be obtained. The bandwidth and capacity of the bottleneck link The estimate becomes inaccurate. On the other hand, if N is excessively large, the amount of probe packets injected into the network increases, and the possibility of packet loss due to buffer overflow increases, thereby increasing the possibility that band-width estimation itself will not be performed.

In the exemplary embodiment of the present invention, the size selection block determining the size of the packet chirp using the binary search based on the minimum value (Nmin) and the maximum value (Nmax) of the size of a predetermined packet chirp .

Thirdly, when exhaustive probing is performed for a wide range of transmission rates r _I starting at the starting transmission rate to build the probe response curve, the physical capacity of the network path to be measured is large, If the utilization of the network path is low, a very large amount of probe traffic must be injected into the network and the probing time becomes long. In the embodiment of the present invention, by using the probe start transmission rate and the packet chord size N corresponding to the automatically selected path characteristics, it is possible to precisely measure the available bandwidth A _t and the capacity C _t of the bottleneck link It provides an automated algorithm to estimate.

2 is a diagram illustrating an automatic band-width estimation method operating in accordance with a traffic characteristic of a network path in a communication system according to an embodiment of the present invention.

Referring to FIG. 2, the automatic band-width estimation method of the communication system according to an embodiment of the present invention determines a probing start transmission rate r _I and a size N of a probe packet checker adaptively to a traffic characteristic of a path A probing parameter selection process 212, an available bandwidth probing process 214 for estimating the available bandwidth A _t by finding the first breakpoint of the probe response curve without performing a full inspection using the selected probing parameters, And a capacity extraction 216 for detecting a capacity C _t using points on the second linear segment obtained in the usable bandwidth estimation process 214.

The probing parameter selection process 212 will be described with reference to FIG. 3, and the available bandwidth probing process 214 will be described with reference to FIGS. 4 and 5, and a capacity extraction process 216 Will be described in Fig.

FIG. 3 is a flowchart illustrating an operation of a probing parameter selection process for determining the probing start transmission rate r _I and the size N of the probe packet checker in accordance with an embodiment of the present invention.

Referring to FIG. 3, the system determines a probing start rate r _I in step 312. The block for determining the probing start transmission rate transmits one packet pilot composed of a plurality of probe packets back-to-back. And then receives the reception rate of the packets transmitted from the side receiving the packet challenge. Then, the system calculates an average reception rate of the packets in the packet retransmission packet and determines the value as the probing start transmission rate. For example, after receiving the arrival time of the first and last packets from the receiving side, the system calculates the number of packets (packet number * packet size) / (last packet arrival time - first packet arrival time) The speed r _I can be determined.

 In general, the reception rate obtained by transmitting a large number of packets in back-to-back is not accurate, but it is easy to use as the probing start transmission rate since it is the value around the usable bandwidth of the bottleneck link. Moreover, unlike the prior art methods, the automated algorithm of the present invention estimates the available bandwidth through iterative probing based on binary search instead of building the probing response curve, so that the bandwidth estimation There is no significant impact on.

After determining the start transmission rate r _I in step 312, the system sets the size N of packet collisions according to the state of the network path while performing steps 314 to 330. The packet chaining size selection block according to the embodiment of the present invention determines the size of the optimal packet chord size through the iterative probing based on the binary search between the minimum value Nmin and the maximum value Nmax of the predetermined packet chirp rate do. In step 314, the system initializes ρ _old to 0, initializes a low value to Nmin, and initializes a high value to Nmax. Wherein ρ is the transmission rate / transmission speed (r _I / _O r), wherein ρ _old may be a ρ value calculated in the previous probing step. Also, the Nmin may be a predetermined minimum packet chirality, and Nmax may be a predetermined maximum packet chirp. In the embodiment of the present invention, Nmin is assumed to be 60, and Nmax is assumed to be 3000.

Then, the system in step 318 m = (low + high) / 2 , and consists of packets, transmitting the configured packet chure is in step 320 the transmission rate r _I (send a packet train of m probe packets with rate r _I )do. And wherein the system receives a rate r _O is received at the receiving side in step 320, they then receive the average speed r _O calculated average reception rate r _O, compute ρ = ρ (representing the transmission rate / transmission speed r _I / r _O ). Then, the system calculates a metric epsilon in step 322. The metric epsilon may be obtained by subtracting p _old from p calculated in step 322 and dividing the metric by p. In step 324, the system compares the metric < RTI ID = 0.0 > e < / RTI > Here, when the metric epsilon is greater than the reference value sigma, it means that the size of the packet collusion is smaller than the proper size, and the fact that the metric epsilon is not larger than the reference value sigma means that the size of the packet collider is larger than the proper size. Here, the reference value? Is a value for adjusting the size of the packet collision, and may be set to 0.05 or less. In the embodiment of the present invention, the reference value? Is assumed to be 0.02.

Therefore, if the metric epsilon is greater than the reference value? In step 324, the system senses the metric in step 324 and increases the low value to the middle point position value in step 326. If the metric epsilon is not greater than the reference value? In step 324, the system reduces the high value to the middle point position value in step 328. After performing steps 326 and 328, the system determines in step 316 whether the value obtained by subtracting the low value from the high value is less than 2 (high-low <2). If the difference is equal to or greater than 2, the process proceeds to step 318 and the above operation is repeated. If the difference between the high value and the low value is less than 1, the system detects this in step 316 and selects one of the high or low values in step 330, Setting. In the embodiment of the present invention, a high value is selected as in step 330 of FIG.

For example, supposing that Nmin is 60 and Nmax is 3000, m is 1530 ((3000 + 60) / 2) in the first 318st step. In this case, the system configures the packet challenge in 1530 packets, and transmits it to the start transmission rate r _I selected in step 312. In step 322, the system calculates a metric epsilon. In step 324, the system compares the metric e with a predetermined reference value sigma. If the metric? Is greater than the reference value? In step 324, the system increases the low value to 1530, which is the middle point position value. In this case, the low value is 1530 and the high value is 3000. Then, the system adjusts the value of m again in step 318 ((3000 + 1530) / 2), and repeats the above process. If the metric? Is not greater than the reference value? In step 324, the system reduces the high value to 1530, which is the middle point position value, in step 328. In this case, the low value is 0 and the high value is 1530. Then, the system adjusts the value of m again in step 318 ((1830 + 0) / 2), and repeats the above process. The above process is repeated when high-low <2.

을 계산한다 (여기서, ρ_old는 이전 프로빙 스텝에서 계산한 ρ 값). 만약 매트릭 ε이 기 설정된 값 σ (예를 들어, 0.02) 보다 크면 low 값을 증가시키고, 아니면 high 값을 감소시켜 m 을 업데이트한 다음에 프로빙을 계속한다. 이 과정을 high-low <2 가 될 때까지 계속하여 패킷추레인 크기 N을 결정한다.As described above, the method of setting the size of the packet chirp of the present invention determines whether the size of the packet chirp obtained for each iteration of the binary search probing is proper by checking whether or not ρ = r _I / r _O converges to a specific value do. In other words, an average reception rate r _O is obtained by transmitting a packet chord comprising m = (low + high) / 2 packets at a transmission rate r _I ,

(Where ρ _old is the ρ value calculated in the previous probing step). If the metric ε is greater than a predetermined value σ (eg, 0.02), increase the low value, or decrease the high value to update m and then continue probing. This process continues until high-low <2 to determine the size N of the packet chirp.

After determining the probing start transmission rate r _I and the packet chaining size N that match the network path characteristics as described above, the process of estimating the available bandwidth of the bottleneck link is started. 4 is a flow chart illustrating an estimation procedure of the available bandwis A _t of a Bartlett link according to an embodiment of the present invention. The present invention uses an iterative probing-based search instead of building the overall probe response curve to effectively find the available bandwis A _t .

Referring to FIG. 4, in step 312, the system initializes variables W _L and W _H for indicating the range of available bandwidth to 0 and initializes input and output rate pairs, Is initialized to zero. After initializing the necessary parameters, the system transmits K packet fragments consisting of N packets using the initial transmission rate r _I in step 414, receives the reception rate r _O from the receiver, It measures the average reception rate r _O of each packet chure (K send probe packet trains of N packets with rate r _I, and compute average output rate r _O of probe packets in each train). Then, in step 416, the probing result evaluation block directly compares each output rate r _O (i.e., reception rate) of K packet fragments with the input rate r _I (i.e., transmission rate) Determine if r _I is greater than the available bandwidth A _t of the bottleneck link (ηK probe trains are asserted to r _I <r _O ?). That is, the probing result evaluation block (block 416) of the system determines that the current probe packet transmission rate is smaller than A _t if the output rate r ₀ of the eta K packet combiner is larger than the transmission rate r _I , And advances the process to the increment block (block 418 to block 420). Conversely, if the output rate of the eta K packet combiner is less than r _I , the probing result evaluation block of the system determines that the current probe packet transmission rate is greater than A _t , and the process proceeds to steps 430 through 436 Block to be executed).

In step 416, if the output rate of the eta K packets is greater than r _I , the system updates the lower limit W _L of the available bandwidth to the probing speed just used in step 418 (W _L = r _I In step 420, r _I is increased by r _I = (W _H + W _L ) / 2, and then the process moves to the next probing decision block in step 422. However, if in the 416 step is ηK packets chure the output speed of less than _{r I (ηK probe trains are asserted} to r I> r O), the system comprising: an upper limit of W _H of the available bandwidth range in step 430. 0 Greater than that. If W _H > 0, which is the upper limit of the available bandwidth, the system stores (x _j , y _j ) = (r _I , r _I / r _O ) using the input and output speed pairs in step 432 record input and output rate pair (x _j, y _j) = (r _I, r _I / r _O)), thereby updating the W _H in 434 steps to r _I, and increase one or j variable (W _H = r _I , j = j + 1). However, if W _H is not greater than zero, the system proceeds to step 434 to update only W _H with r _I (W _H = r _I , j = j + 1) without storing the input and output speed pairs. Thereafter, the system moves as _I = r (W _H W + _L) / 2, then reducing r _I a, and then probing the decision block of step 422 (next probing decision) in step 436.

The probing decision block of the system compares the size of the band-wise range W _{H -} W _L with a predetermined threshold value ω. That is, in step 422, the system compares the value obtained by subtracting the lower-limit band-wise value W _L from the upper-limit band-wise value W _H with the set reference value ω (W _H - W _L ≤ ω). At this time, if the size (W _H -W _L ) of the band-wise range is smaller than?, The system detects this in step 422 and stops probing in step 440 and obtains A _t = (W _L + W _H ) / 2 Output to the available bandwith of the bottle neck path. However, if the size (W _H - W _L ) of the band-wise range is equal to or greater than?, The system returns to step 414 and repeats the above operation with a new input speed r _I.

In FIG. 4, it is assumed that K is 12 and eta is 0.6. In this case,? K is 7.2. At this time, in step 414, K (= 12) packet chirps comprising N packets are transmitted using the initial transmission rate r _I , and an average reception speed r _O of each packet chirp is measured. And directly compares each output rate r _O (ie, reception rate) of the K (= 12) packet combiner with the input rate r _I (ie, transmission rate) to determine the current probe packet input rate r _I Gt; A _{&lt; /} RTI _&gt; is greater than the available bandwidth A _t .

At this time, if the probing result evaluation is that the output rate r ₀ of ηK (= 7.2, here may be 8 packet chirps) or more packet colliders is larger than the transmission rate r _I , it is determined that the current probe packet transmission rate is smaller than A _t do. Then, the system updates the lower limit W _L of the available bandwidth in step 418 to the probing speed just used for probing (W _L = r _I ), and in step 420, r _I = (W _H + W _L ) / 2 To increase r _I. Then, the system compares the size of the band-wise range with a predetermined threshold value (W _H - W _L ??). At this time, if the size (W _H - W _L ) of the band-wise range is smaller than?, The system stops probing in step 440 and adds A _t = (W _L + W _H ) / 2 to the available band . However, if the size (W _H - W _L ) of the band-wise range is equal to or greater than?, The system returns to step 414 and repeats the above operation with a new input speed r _I. By repeating the above operation, the transmission speed r _I is increased, so that the lower limit range W _L of the band wing is close to the available band width range A _t .

Also, if the probing result evaluation is less than r _I , the output rate of the ηK (= 7.2, here may be eight packet chirp) packet resolvers is less than r _I , then the system determines that the current probe packet transmission rate is greater than A _t In step 430, it is determined whether W _H , the upper limit of the available bandwidth, is greater than zero. If W _H > 0, which is the upper limit of the available bandwidth, the system stores (x _j , y _j ) = (r _I , r _I / r _O ) using the input and output rate pairs in step 432, In step 434, W _H is updated to r _I. However, if W _H is not greater than 0, the system proceeds to step 434 to update only W _H with r _I without storing the input and output speed pair. Then, the system performs After reduction, the probing operation determined by the r _I = r _{I (H} W _L + W) / 2 at step 436. The probing decision operation is as described above. By repeating the above operation, the transmission rate r _I is reduced, and the upper limit range W _H of the band wise becomes close to the usable band width range A _t .

Therefore, according to the probing result evaluation in step 416, the lower and upper limits W _L and W _H of the band-wise range are approximated to the available band-wise A _t , It becomes possible to estimate the usable band-wise A _t . (X _j , y _j ) = (r _I , r _I / r) for each of the W _Hs (updated transmission rates r _I ) updated in the process of probing the upper limit W _H of the usable band- _O ), which is used to estimate the capacity C _t at the back end. That is, the change ((x _j , y _j ) = (r _I , r _I / r _O ) of the upper limit W _H of the usable band width range corresponds to the information between A _t and B in the response curve F of FIG. do.

5 is a flow chart illustrating an estimation procedure of the available bandwis A _t of a bottleneck link according to another embodiment of the present invention.

Referring to FIG. 5, in the process of estimating the available bandwidth A _t of the bottleneck link, the transmission rate may be located around the available bandwidth A _t or there may be an area where it is difficult to determine the transmission band. This region is called a gray region. As described above, when K is 12 and eta is 0.6, the eta K is 7.2. At this time, if the probing result evaluation is such that the output rate r ₀ of packet colliders smaller than ηK (= 7.2, here, 8 packet collisions may be) is larger than the transmission rate r _I and the output of packet colliders smaller than ηK speed r ₀ when the transmission rate smaller than the rating _I r, r _I transmission rate, it is difficult to late or faster than the evaluation desired available bandwidth a _t. For example, the six packets chure the output of the speed r ₀ greater than the transmission rate r _I, may also be four packet chure output speed r is ₀ is less than the transmission speed of _I r. In this case, the probing speed is adjusted by setting the neutral region in FIG. That is, FIG. 5 estimates the available bandwidth A _t in consideration of the transmission rate located in the neutral region.

The system first initializes the variables W _L and W _H for indicating the range of the usable bandwith to 0 and initializes the variables G _L and G _H for expressing the range of the neutral region to 0 for probing in step 512. In step 514, the system initializes the necessary parameters, and then, in step 514, the system measures the reception rate r _O of each packet chirp by transmitting K packet fragments consisting of N packets using the initial transmission rate r _I. The probing result evaluation block then directly compares each output rate r _O (i.e., reception rate) of the K packet fragments with the input rate r _I (i.e., transmission rate) to determine whether the current probe packet input rate r _I is greater than the bottleneck Is greater than the available bandwidth A _t of the link. In step 516 and step 520, the probing result evaluation block of the present invention determines that the current probe packet transmission rate is less than A _t if the output rate of the eta K packet checker is greater than r _I , Advances the process to an incremental block. Conversely, if the output rate of the eta K packet replica is less than r _I , the process proceeds to steps 532 through 542 for determining the current probe packet transmission rate to be greater than A _t . However, if both are not included, the process proceeds to the gray region processing block in steps 550 to 560.

In operation 518, the system updates the lower limit W _L of the available bandwidth in step 518 with the probing speed used for probing, and checks the variable G _L in step 520 . Variable G _L> 0 if the system is _{r I = (W H + W} L) / 2 to increase the r _I and a, otherwise the system is r _I = (G _L + W _L) in step 524. In step 522 / 2 increases r _I. On the other hand, the 532 steps - Looking at the operation of the 5542 phase probing rate decrease block, the system checks the W _H is the lower limit of available bandwidth range in step 532, and if the available bandwidth upper limit of W _H in the range of> 0 The system stores (x _j , y _j ) = (r _I , r _I / r _O ) using the input and output speed pairs in step 534, and then updates W _H to r _I in step 536. However, if W _H is not greater than zero, the system updates W _H only to r _I in step 536 without storing the input and output speed pair. Then, the system examines the variable G _H in step 538, a variable G _H> 0 is and _{r I = (W H + W} L) / 2 decreases the r _I a at 540 steps, otherwise r _I = in 542 steps in _{(G H + W H) /} 2 decreases the r _I. After updating the transmission rate value r _I in the probing rate increasing or decreasing block, the system moves to the next probing decision block in step 570.

In addition, if it is determined in the case in the step 516 is ηK packets chure the output rate of no greater than r _I is ηK packets chure the output rate in step 520 is not less than r _I, the system is not part of neither in both, The process proceeds to steps 550 through 560 of a gray region processing block. In this case, it is checked in step 550 whether the lower limit G _L and the upper limit G _H of the gray area range are 0 (G _L = G _H = 0). At this time, if G _L = G _H = 0, the lower limit G _L and the upper limit G _H of the gray area range have never been updated. Therefore, the system first updates G _L and G _H to r _I in step 552, To the block for judging whether G _H < = r _I. On the other hand, if both G _L and G _H are not zero (i.e., if they have been updated at least once), the system immediately goes to a block that determines whether G _H < = r _{I in} step 554. If the decision block in step 554 is true, the system updates G _H to r _I and updates r _I to (W _H + G _H ) / 2 in step 556. Otherwise, if the result of this decision block is false, the system moves to block 558 to determine if G _L > r _I. Where G _L > r _I, the system updates G _L to r _I and updates r _I to (W _L + G _L ) / 2 in step 560. Otherwise, if G _L < = r _I, then the system moves to the next probing decision block, which determines whether to proceed with the next probing with the new r _I.

The system compares the size (W _H - W _L ) of the first banding range with a predetermined threshold value? In the next brozing decision block in step 570. If the size of the band-wise range is less than omega, the system stops probing in step 572 and outputs A _t = (W _L + W _H ) / 2 to the available bandwith of the bottleneck path. Also, even though G _L - W _L and W _H - G _H are not greater than?, The system ends probing in step 572 and returns A _t . However, if these conditions are not satisfied, the system proceeds from step 517 to the probing block in step 514 to repeat the above operation using the new input rate r _I.

6 is a flowchart illustrating a procedure for estimating a capacity of a bottleneck link according to an embodiment of the present invention.

Referring to FIG. 6, after determining the available bandwidth A _t , the system performs a process of detecting the capacity of the bottleneck path. When the capacity estimation procedure is started, the system calculates an available bandwidth A _t , an upper limit value W _H of the band-wise range, and a second linear segment of the probing response curve in step 610 of FIG. A number of points (x _j , y _j ) that are supposed to be input are input. And that the system is to select as the point (x _j, y _j) As is in the first transmission rate is larger than W _H at the same time, the anchor point (anchor point) to the nearest point on the A _t the input in step 612, in step 614 And calculates the slope a _j between the anchor point and all other points satisfying x _j > = W _H.

Then, in step 616, the inverse of the slope obtained in step 616 may be the bottleneck capacity C _t. Therefore, it is determined whether the slope a _j values of the two points obtained are appropriate. That is, the system checks whether the calculated slope a _j value is less than 0 (a _j? 0), or whether the slope a _j value is greater than the available band weight A _t (a _j <1 / A _t ). At this time, if the above condition is satisfied, the system confirms this in step 616, and proceeds to step 618 to discard the corresponding slope a _j value. That is, the physical capacitance sheet of the link can not be a negative number a _j <= 0 values are discarded, and discards FIG soluble bandwidth A _t are not be larger than the physical capacitance City _{C t a j <1 / A} t value. Then, the system outputs in step 620 by the inclination calculating the capacitance City estimation value using a _j and (compute capacity estimate C _t using remaining a _j), neck bottle on the calculated result from the 622 phase capacitance City C _t. That is, in order to increase the accuracy of the capacity estimation value, the system performs steps 616 through 618 while excluding the inappropriate values, which are the products of the measurement errors, and stores the reciprocal of the average value of the remaining values in step 620 as Bottle- .

Figure 7 is an illustration of an example of an architecture for performing the available bandwidth and capacity estimation of a bottleneck link in accordance with an embodiment of the present invention.

Referring to FIG. 7, the PS indicates a transmitting side system for transmitting a probe packet, and the PR is a receiving side system for receiving and responding to a probe packet transmitted from the system PS. A plurality of repeaters may be connected between the system PS and the PR. In FIG. 7, five repeaters R1 to R5 are connected. Therefore, in the case of FIG. 7, it may be a five-hop model. In addition, S1-S4 and D1-D4 are elements that share limited resources on the network and can be factors that directly affect the available bandwidth and capacity of the bottleneck link. Also, a transmission delay may occur in the packets transmitted from the PS system through the repeaters R1-R5.

At this time, the PS system transmits a probe packet pilot with a transmission rate r _I , and the PR system measures a reception rate of a packet pilot transmitted from the PS system and feeds back the result to the PS system. In this case, the system for performing the available bandwidth and capacity estimation procedure of the bottleneck link as shown in FIGS. 2 to 6 may be a PS system.

In FIG. 7, the PS may include a PRC-MT (PRC Meshment Tool) capable of estimating available bandwidth and capacity on an arbitrary multi-hop path. Hereinafter, a procedure for estimating the available bandwidth and the capacity will be described with reference to FIGS. 2 to 6.

First, the process of selecting the parameters r _I and N will be described.

내에 있는 응답 커브(fluid curve) F의 값과 평행하게 되는 값으로 수렴한다(converges). 이는 프로브 패킷 추레인의 길이 N이 커질 때, r_I/r_O의 비율이 특정 값에서 포화되는 것으로 해석될 수 있다. 이를 확인하기 위해, 다른 N 값을 위한 r_I/r_O의 양(quantity)을 검사한다. 이를 위하여 가변적인 N을 가진 패킷들을 가용 밴드위스의 속도(예를들면, A_t=61Mb/s) 보다 높은 속도(예를들면 68Mb/s(r_I=68Mb/s)로 전송한다. r_I/r_O은 N이 증가함에 따라 응답 범위(fluid-bound)보다 조금 큰 값(예를들면 (λ_t +r_I)/C_t = 1.07)으로 빠르게 줄어든다. 이에 주어진 입력 속도(r_I)에 있어서, r_I/r_O의 비율이 포화되도록 하는 N을 반복적으로 프로빙하는 실증적 방법(empirical method)이 검토된다.For sufficiently large N values, the slope of the second linear segment (the segment between A _t and B in Fig. 1)

Converges to a value that is in parallel with the value of the fluid curve F in the fluid. This can be interpreted as the ratio of r _I / r _O saturates to a certain value when the length N of the probe packet probe is increased. To confirm this, we check the quantity of r _I / r _O for another N value. Variable N packets the speed of the available bandwidth with For this purpose (e.g., A _t = 61Mb / s) rate higher than (e.g., sent to _{68Mb / s (r I = 68Mb} / s). R I / r _O is a little larger than the response range (fluid-bound) (for example, _{(λ t + r I) /} C t = 1.07) to decrease quickly. in any given input speed (r _I), as N increases , An empirical method of repeatedly probing N such that the ratio of r _I / r _O saturates is examined.

As described above, the embodiment of the present invention proposes a simple N value selection procedure for adjusting N based on variation of r _I / r _O at the transmission rate r _I. Although the input transmission rate of such a routine is not specifically limited here, since the change in the r _I / r _O ratio at an input rate smaller than A _t is rapidly reduced even with a small increase in the length N of the probe chain, It is preferable to use a speed that is not too smaller than the available bandwidth A _t of the path. For this purpose, it is sufficient to find a velocity not less than A _t , and the asymptotic dispersion rate (ADR) of the path is proved to be greater than A _t , so ADR is used as the initial value of r _I Suitable. Then, the PRC-MT transmits a single packet header to probe the ADR and calculates ADR = q / E [y] at the receiver (E [y] (inter-packet dispersion), q is the probe packet size, and N is the probe-train length. After the input transmission rate r _I is determined, the remaining procedures for probing the chirp length are as follows.

ρ is defined as the ratio of the current input and output speeds (ρ = r _I / r _O ), and ρ _old is defined as the previous value of ρ. Further, Nmin and Nmax are respectively defined as a minimum value and a maximum value of a chirp length which can be adjusted by the user, and [sigma] is defined as a specific threshold variable which can vary between 0 and 1. The selection routine selects Nmin and Nmax And performs a binary search between them. In this procedure it is tested whether the ρ is converging to a specific value for a given N length by sending an N length packet chirp with the same r _I rate (r _I = ADR) as the ADR to the receiving end. To determine the value at which ρ is saturated, the relative error metric ε (ε = | ρ-ρ _old | / ρ) described below is calculated in the selection procedure. Thus, the routine reduces the N value if? Is less than or equal to?, Otherwise increases the N value.

Users can use any packet size if qmin≤q≤qmax (qmin = 200bytes; qmax = 1500bytes is used). In this case, Nmin and Nmax are scaled-up or scaled-down by qmax / q. This routine ensures that when a user selects a smaller packet q, a larger N value is selected to satisfy the condition to stabilize the fluctuation of the linear segments of the response curve F. In the embodiment of the present invention, it is assumed that Nmin = 60, Nmax = 3000,? = 0.02, and q = 200 bytes.

Second, we examine the bandwis probing process (the process of estimating A _t and C _t ).

Let's look at how to identify the first breakpoint of the response curve F, which is the point at which the input transmission rate r _I is greater than the arrival rate r _{O with a} probe chirp length N. To efficiently retrieve this point, the PRC- MT utilizes a repetitive search based on probing similar to Pathload. On the one hand, these two tools differ in how the input speed r _I corresponds to A _t . For example, PRC-MT directly compares r _I and r _O to determine whether the current rate r _I is greater than A _t while the pass load examines one-way delays of probe packets, It is estimated whether the speed r _I is greater than A _t .

In probing, the PRC-MT sends a group of K packet routers at a given rate r _I. Then, the transmission rate r _I of the next K probe probes is adjusted depending on how much of the packet chirality η falls within the range of r _I > r _O or r _I <r _O. More specifically, if PRC-MT indicates that the velocity r _I of the? If it is determined to be greater than r _O (r _{I> O} r), a smaller r _I, and, if their speed is determined to be smaller than r _I r _O (r _{I <O} r), and increasing the r _I. If neither of these cases is true (ie, the number of probe probes belonging to r _I > r _O or r _I <r _O is less than ηK), pass-loading may result in a gray region ).

를 타이트 링크(tight link)의 가용 밴드위스 추정 값으로 리턴 한다. PRC-MT에서는 디폴트 값(default value)으로 η=60% 또한 K=12로 가정한다.[W _L, W _H ] is the available band-width range updated after each K probe chirp measurement round. W _L represents the highest speed among the speeds identified as lower than A _t in a particular round and W _H represents the lowest one of the speeds identified above the A _t by the round. After each probing round, the PRC-MT updates the band-wise range and selects a new probing speed r _I for the next round using a pass-through approach. This search process is based on the fact that the band-wise range [W _L, W _H ] around A _t is less than the measured ADR or a specific threshold value ω (ie, ω = 0.02 ADR) that can be automatically selected based on the value given by the user It continues until it becomes small. The PRC-MT, when its internal algorithm terminates,

As an estimated usable bandwidth value of the tight link. In the PRC-MT, the default value is assumed to be η = 60% and K = 12.

을 산출하기 위해 사용될 수 있다.After probing of the available bandwith, the PRC-MT begins to estimate the tight link capacity _Ct . The main focus of this routine is to select two points to be utilized to infer C _t in the second linear segment. In order to facilitate the process of estimating C _t , the PRC-MT determines the transmission rate r _I and its counterpart each time r _I is greater than r _O, while r _I decreases the upper limit W _H , while probing the available bandwidth. Lt; / _RTI &gt; The points thus recorded are the estimated values of the capacities

. &Lt; / RTI &gt;

In the above cases (i.e., when the second segment is perfectly straight and there is no measurement noise), any two points among the recorded points to infer C _t may be selected. However, there is no simple way to select the optimal two points with a specific measurement noise and an incomplete straight line. The following empirical method is required.

의 조건을 만족시키는 포인트를 첫 번째 포인트로 사용하지 않는 두 가지 이유가 있다. 첫 번째 이유는 응답커브 F에서 두 번째 선형 세그먼트가 A_t 주변에서 시작하는 지점이 매우 명확하지 않기 때문이다(도 2(a) 참조). 다른 이유는 추정된 값

의 측정 에러(예를 들어,

)에 의해 상기 포인트가 첫 번째 선형 세그먼트 상에 있을 수 있기 때문이다. 이러한 경우, 캐패시티를 측정하는데 있어서 많은 에러를 초래 할 수 있다. x _i and y _i (i = 1, ..., m) have two or more recorded points (x ₁ , y ₁ ), ..., (x _m , y _m ) Lt; RTI ID = 0.0 &gt; transmission / reception &lt; / RTI &gt; PRC-MT is first to select a (x _i, y _i), where x _i is the smallest value among the m number of points satisfying the conditions of x _i ≥W _H. Meanwhile,

There are two reasons for not using a point that satisfies the condition of the first point. The first reason is that the point at which the second linear segment starts around A _t in response curve F is not very clear (see Fig. 2 (a)). Another reason is that the estimated value

0.0 &gt; (e. G., &Lt; / RTI &

), The point may be on the first linear segment. In such a case, it may cause a lot of errors in measuring the capacity.

를 만들기 위해서는 두 번째 포인트가 두 번째 선형 세그먼트 상에 있어야 한다. PRC-MT가 기록된 포인트들 중 전송 속도 x_i가 W_H에 가장 가까운 첫 번째 포인트(x_i, y_i)를 고르기 때문에, 두 번째 선형 세그먼트 상에 있을 확률이 높은 첫 번째 포인트에서 가까운 포인트를 두 번째 포인트로 고려할 수 있다. 그러나 만약 두 개의 포인트들이 서로 가까이 위치한다면, 이 두 포인트들의 선형 기울기를 산출하는 것은 측정 노이즈에 더 많이 노출 될 수 있다. 이에, 두 번째 포인트는 두 번째 선형 세그먼트 상에 있기만 한다면, 첫 번째 포인트로부터 될 수 있는 한 멀리 떨어진 포인트 인 것이 바람직하다. 이에, C_t를 추정하기 위해 기록된 포인트들 중 가정 멀리 떨어진 두 포인트들이 선택된다. 상기 서술한 바에 근거하여, PRC-MT는 (x_j, y_j)를 두 번째 포인트로 사용한다. 여기서 x_j는 기록된 포인트들 중 가장 큰 값이다. Accurate Capacity Estimation

The second point must be on the second linear segment. Because the PRC-MT selects the first point (x _i , y _i ) closest to W _H , the transmission point rate x _i is the point closest to the first point that is likely to be on the second linear segment It can be considered as a second point. However, if two points are located close to each other, calculating the linear slope of these two points may be more exposed to the measurement noise. Thus, the second point is preferably a point as far as possible from the first point, as long as it is on the second linear segment. Thus, two far-lying points of the recorded points are selected to estimate C _t . Based on the above description, the PRC-MT uses (x _j , y _j ) as the second point. Where x _j is the largest of the recorded points.

를 산출한다.Between the two points (x _i, y _i) and (x _j, y _j) to select and then, PRC-MT is _{(x i, x i / y} i) and _{(x j, x j / y} j) linear segment The estimated value of the tight link capacity, which is the reciprocal of the slope of

.

If the number of recorded points m is less than two, the PRC-MT sends additional packets with velocities r _I greater than A _t to obtain (r _I , r _O ) pairs. Such a case rarely occurs only when a very large threshold value ω is used to terminate the algorithm.

Although not shown in the drawings, in an apparatus for estimating the available bandwidth A _t and capacity C _t of network paths in a communication system according to an embodiment of the present invention, an apparatus for estimating available bandwidth A _t and capacity C _t is provided between a minimum value Nmin and a maximum value Nmax A parameter selection unit for determining an optimal packet chore size N through iterative probing based on a binary search and a transmission rate r _I and a reception rate r _O of the determined plurality of K packet colliders, An available bandwidth estimating unit that estimates an available bandwidth while repeating an operation of increasing the probing speed and decreasing the probing speed and storing the point when the receiving rate is low; A point close to the band wise is determined as an anchor point, and an inclination between the anchor point and the remaining points is obtained, And a capacity estimating unit for estimating the null capacity.

Here, the parameter selection unit may include a start transmission rate selection unit and a packet crosstalk size determination unit, and the start transmission rate selection unit may select one of the plurality of probe packets as a back-to-back And calculating the average reception rate of the packets in the packet retransmission packet and determining the value as the probing start transmission rate.

The packet chunk size determining unit calculates a current transmission rate / average output rate after determining the average reception rate, and transmits the current transmission rate / average rate and the previous transmission rate / If the error matrix is larger than the set value, the intermediate value of the upper and lower limit values of the packet size is set as the lower limit value. Otherwise, the intermediate value of the upper and lower limit values of the packet size And repeats the above operation. In the iterative process, the size of the packet chirp with the difference of the upper and lower limit values of the packet chirality of 1 or less can be set to an optimal packet chirpy size.

The available band-width estimator may further include: a transmitter for transmitting a plurality of K packet colliders at an initial transmission rate r _I ; a receiver for calculating an average reception rate r _O of the packet colliders; comparing the transmission rate and the receiving rate (ηK probe-trains are asserted to r I <r O or r I> r O), the available band reception rate is higher transmission rate than the transmission rate of the plurality of packets chure Switzerland A probing result evaluating unit that determines that the transmission rate is greater than the available bandwidth if the reception rate of the plurality of packet colliders is smaller than the transmission rate; set of the reception rate lower limit value W _L of the greater bandwidth than the transmission rate to the transmission rate, and transmitting the upper and middle value of the lower limit value _{(r I = (W H +} W L) / 2) of the bandwidth At speed A probing speed increasing unit for increasing a probing speed of the plurality of packet collisions by the probing result evaluating unit and storing the current point position and setting the upper limit value W _H of the bandwidth to the transmission rate, A probing speed decreasing unit that sets the intermediate value (r _I = (W _H + W _L ) / 2) of the upper and lower limit values of the band wise as a transmission rate, and a difference between an upper limit and a lower limit value of the band- _H W _L ??), The middle of the upper and lower limit values of the band wise is determined as an available band width (A _t = (W _H + W _L ) / 2) And a probing decision unit.

The capacity estimator may include an anchor point determination unit for inputting the available bandwidth, the upper limit value of the band wise and the information of the pointers from the available bandwidth estimation unit and determining a point close to the available bandwidth as an anchor point, A slope calculating unit for calculating a slope between a point and all remaining points, and an estimating unit for obtaining an average value of the slopes and estimating an inverse number of the average value with a Barthelem capacity.

The embodiments of the present invention disclosed in the present specification and drawings are merely illustrative of specific embodiments of the present invention and are not intended to limit the scope of the present invention in order to facilitate understanding of the present invention. It will be apparent to those skilled in the art that other modifications based on the technical idea of the present invention are possible in addition to the embodiments disclosed herein.

Claims (10)

A method for estimating available bandwidth and capacity of network paths in a communication system,
A parameter selecting step of determining a size of a packet chirp through repeated probing based on a binary search between a minimum value and a maximum value of a size of a packet chirp,
If the reception rate is higher than the transmission rate, and if the reception rate is lower than the transmission rate, comparing the transmission rate and the reception rate of a plurality of packet colliders having the determined packet chirality, Estimating an available bandwidth while repeating the operation of storing the point;
Determining a point at which the transmission speed is close to the available bandwidth as an anchor point after estimating the available bandwidth and estimating bit rate capacity by obtaining a slope between the anchor point and the remaining points; Gt; 2. The method of claim 1, wherein the parameter selection process is performed by back-to-back transmission of a packet packet comprising a plurality of probe packets, and then calculates an average reception rate of the packets in the packet packet, Value as a probing start transmission rate. 3. The method of claim 2, wherein the step of determining the size of the packet chirps comprises:
Calculating a current transmission rate / average output rate after determining an average reception rate, transmitting the packet transmission rate at the transmission rate,
Calculating an error matrix using the current transmission rate / average output rate and the previous transmission rate / average output rate;
If the error matrix is larger than the set value, the medium size of the upper and lower limit values of the packet header is set as the lower limit value, otherwise, the medium size of the upper and lower packet values of the packet header is set as the upper limit value, Process,
And setting a packet chirp at which the difference between the upper and lower limit values of the packet chirp is less than or equal to 1 in the iterative process as an optimal packet chirp. 4. The method of claim 3, wherein the step of estimating the available band-
Transmitting the plurality of packet colliders at an initial transmission rate and calculating an average reception rate of the plurality of packet colliders,
And determines that the transmission speed is smaller than the available bandwidth if the reception speed of the plurality of packet colliders is greater than the transmission speed, and determines a reception speed of the plurality of packet colliders If the transmission rate is smaller than the transmission rate, it is determined that the transmission rate is greater than the available bandwidth;
And setting a lower limit value of the bandwidth to the transmission rate if the reception speed of the plurality of packet colliders is greater than a transmission rate in the evaluation of the probing result and probing that the intermediate value of the upper and lower limit values of the band- The speed increase process,
If the reception rate of the plurality of packet colliders is less than the transmission rate in the evaluation of the probing result, sets the upper limit value of the band wise to the transmission rate, stores the current point position, A probing speed reduction process of setting the intermediate value as a transmission rate,
If the difference between the upper and lower limit values of the band wise is within a reference value, determining an intermediate band width between the upper and lower limit values of the band wise as an available band wise. 5. The method of claim 4, wherein the step of estimating the capacity comprises:
Determining an anchor point as a point nearest to the usable bandwith;
Calculating an inclination between the anchor point and all remaining points;
Calculating an average value of the slopes, and estimating an inverse number of the average value using a Barthelem capacity. 1. An apparatus for estimating available bandwidth and capacity of network paths in a communication system,
A parameter selector for determining a size of a packet chirp through repeated probing based on a binary search between a minimum value and a maximum value of a size of each of the predetermined packet chirps,
If the reception rate is higher than the transmission rate, and if the reception rate is lower than the transmission rate, comparing the transmission rate and the reception rate of a plurality of packet colliders having the determined packet chirality, An available bandwidth estimating section for estimating an available bandwidth while repeating the operation of storing the point,
And a capacity estimator for estimating the bit rate capacity by determining a point at which the transmission speed is close to the available bandwidth after the estimation of the usable bandwidth range as an anchor point and a slope between the anchor point and the remaining points, . 7. The apparatus of claim 6,
After one packet of the plurality of probe packets is transmitted back-to-back, the average reception rate of the packets in the packet is calculated, and the probe start rate is determined as the probe start transmission rate Further comprising a transmission rate selection unit. The apparatus of claim 7, further comprising: a parameter selector for determining the size of the packet chirp;
Transmits a packet header at the transmission rate, calculates an average reception rate, calculates a current transmission rate / average output rate,
An error matrix is obtained using the current transmission rate / average output rate and the previous transmission rate / average output rate,
If the error matrix is larger than the set value, sets the intermediate size of the upper and lower limit values of the packet summation as the lower limit value, otherwise sets the middle size of the upper and lower limit values of the packet summation as the upper limit value,
And repeating the above-described operations, wherein in the iterative process, a packet collision factor in which the difference between the upper and lower values of the packet collision factor is equal to or less than 1 is set as an optimal packet collision factor. 9. The apparatus of claim 8, wherein the available band-
A transmitter for transmitting the plurality of packet colliders at an initial transmission rate;
A receiving unit for receiving an average rate of reception of the plurality of packet colliders,
And determines that the transmission speed is smaller than the available bandwidth if the reception speed of the plurality of packet colliders is greater than the transmission speed, and determines a reception speed of the plurality of packet colliders A probing result evaluating unit that determines that the transmission rate is larger than the available bandwidth if the transmission rate is smaller than the transmission rate,
Wherein the probing result evaluating unit sets the lower limit value of the bandwidth to the transmission rate if the reception speed of the plurality of packet colliders is larger than the transmission rate and sets the intermediate value of the upper and lower limit values of the band- A speed increasing portion,
Wherein the probing result evaluating unit stores the current point position and sets the upper limit value of the band wise to the transmission rate if the reception rate of the plurality of packet colliders is smaller than the transmission rate, A probing speed decreasing unit that sets the intermediate value as a transmission rate,
And a probing decision unit for determining an intermediate band width between the upper and lower limit values of the band wise as an available band width when the difference between the upper and lower limit values of the band wise is within a reference value. 10. The apparatus of claim 9,
An anchor point decision unit for inputting information on the available bandwidth, the upper limit of the band-wise value and the pointers from the available bandwidth estimation unit and determining a point nearest to the available bandwidth as an anchor point;
A slope calculating unit for calculating a slope between the anchor point and all remaining points;
And an estimator estimating an average value of the slopes and estimating an inverse number of the average value by a Barthelem capacity.

Images