JP2012175301A - Band calculation method and device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To calculate a band needed for realizing a targeted throughput in a network which accommodates communication terminals having a flow control function to control a transmit rate by autonomous window control and performs communication by a packet exchange.SOLUTION: The present invention includes means of presetting, for each flow accommodated in a network, a targeted throughput every characteristics of the respective flows, and of calculating, as a necessary band, a band whose cumulative probability in the probability distribution of a band used during simultaneous connections as will be needed when all flows accommodated in the network maintain their targeted throughput is equal to or less than a prescribed value m.

Description

  The present invention relates to a communication quality control technique in a packet communication network such as the Internet or an intranet, and more particularly to a bandwidth calculation method and apparatus for calculating a required bandwidth in a network accommodating a plurality of TCPs (Transmission Control Protocol).

  In recent years, broadband communication networks have been increasing, and more traffic has been distributed over the Internet. In addition, it has been reported that most of the Internet traffic is TCP communication (for example, see Non-Patent Document 1), and high-quality communication such as VPN and video distribution is required as the network infrastructure becomes wider. An example of realizing services with TCP has also appeared.

  On the other hand, TCP is a protocol having an autonomous transmission rate control mechanism by window control, and is greatly affected by traffic existing on the same route, particularly other flows. Therefore, the throughput achieved by the interference between flows varies greatly. Therefore, the network link bandwidth required for achieving the target throughput for each flow generated in the network is not easily obtained, and has only to be set empirically.

  On the other hand, there is a theoretical formula for analytically deriving the throughput of a flow under a given environmental condition that targets throughput calculation in a “network environment in which only one flow exists in a steady state” (for example, , See Non-Patent Document 2)

K. Cho, K. Fukuda, H. Esaki and Akira Kato, "Observing Slow Crustal Movement in Residential User Traffic," ACM CoNEXT2008, Dec. 2008. Jitendra Padhye, Victor Froiu, Don Towsley, Jim Kurose, "Modeling TCP Throughput: A Simple Model and its Empirical Validation", ACM-SIGCOMM, pp.303-314, Sep. 1998.

  The theoretical formula of Non-Patent Document 2 described above is intended for throughput calculation in a “network environment in which only one flow exists in a steady state”, and is different from the flow throughput in an actual network environment. For example, it is known that when the flow size of the flow is extremely small, the throughput is much smaller than the throughput represented by the theoretical formula.

  Further, since a normal network accommodates a plurality of flows, interference occurs between the flows, and the flow throughput changes. In addition, the target throughput for each flow is rarely the theoretical maximum value in the steady state, and the required bandwidth is calculated by directly applying the theoretical formula as a bandwidth calculation means for the network that accommodates the flow. In this case, the necessary bandwidth is excessively calculated.

  For this reason, it is difficult to say that the target value of the flow analytically calculated by the method of Non-Patent Document 2 is the actual throughput that the flow actually targets in the original network environment.

  The present invention has been made in view of the above points, and by predetermining a target throughput for each flow accommodated in the network, it is possible to calculate a bandwidth necessary for realizing the target throughput. An object of the present invention is to provide a simple band calculation method and apparatus.

In order to solve the above-described problem, the present invention provides a reception function having a function of returning a reception confirmation packet to the transmission terminal when the reception terminal receives a data packet from the transmission terminal via a network that performs communication by packet switching. A transmission terminal having a function of grasping the amount of data received by the reception terminal by a reception confirmation packet, and a communication system in which a window size increases every time a reception confirmation packet from the reception terminal arrives A bandwidth calculation device in a network to accommodate,
Throughput calculation means for setting a target throughput for each flow characteristic for each flow accommodated in the network;
Bandwidth calculating means for calculating a bandwidth in which the cumulative probability of the target throughput total value of each flow accommodated in the network is less than or equal to a certain value.

  As described above, according to the present invention, in a network accommodating a plurality of flows, it is necessary to realize a target throughput set for each flow by predetermining a target throughput for each flow accommodated in the network. Bandwidth can be calculated.

It is a block diagram of the zone | band calculation apparatus in the 1st Embodiment of this invention. It is a block diagram of the zone | band calculation apparatus in the 2nd Embodiment of this invention. It is a flowchart of the throughput calculation process in the 2nd Embodiment of this invention. It is a flowchart of operation | movement of the zone | band calculation part of 1st Example of this invention. It is an example of the time change of the flow simultaneous connection zone | band of 1st Example of this invention. It is an example of the complementary distribution of the accumulation probability distribution of the required zone | band of 1st Example of this invention. It is the relationship between the flow size of the 2nd Example of this invention, and the estimated upper limit of throughput. It is an example of the time change of the simultaneous connection zone | band of the flow of 2nd Example of this invention. It is an example of the complementary distribution of the cumulative probability distribution of the required band of 2nd Example of this invention. It is a comparison with the estimated value of the throughput upper limit of the 2nd Example of this invention, and a simulation result.

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

  The present invention is directed to a network that performs communication by packet switching. The network accommodates communication terminals having a flow control function for controlling the transmission rate by autonomous window control.

[First embodiment]
FIG. 1 shows the configuration of a bandwidth calculation apparatus according to the first embodiment of the present invention.

  The bandwidth calculation apparatus shown in FIG. 1 includes a throughput calculation unit 10 and a necessary bandwidth calculation unit 20.

The throughput calculation unit 10 sets a target throughput for each flow characteristic in advance for each flow accommodated in the network. Examples of the flow characteristics include, but are not particularly limited to, the file size of the flow, the application type, the distance to the communication partner, and the like. The target throughput to be set may be a fixed value such as 100 kbps per flow, or may be a different value for each file size, a value calculated from the flow connection time, or the like, but is not particularly limited in the present embodiment. Further, the method for determining throughput is not particularly limited.
・ Because it is HD video distribution, set it to 15Mbps;
・ Since the video quality is SD, 8Mbps
-Since the file size is XMB, set the throughput to (X * 8) / (Y * 60) Mbps to keep the communication time within Y minutes;
There are various methods.

  Hereinafter, a method for calculating the target throughput of each flow in the present embodiment will be described.

  The throughput calculation unit 10 calculates the target throughput of each flow as α times the throughput obtained as the maximum throughput that can be realized in a state where there is no interference between communications.

  Specifically, a process for calculating a realizable maximum throughput value will be described.

  TCP's transmission rate is controlled by autonomous window control, and its throughput changes dynamically according to the communication status. However, the bandwidth and buffer size on the network access side, round-trip propagation delay excluding queuing delay, etc. Is determined, the maximum throughput that can be achieved when there is no interference between the flows. In particular, the number of packets exchanged in the flow is determined according to the file size of the flow, and the transition of the congestion window size is determined. Therefore, the maximum throughput that can be realized is determined according to the flow size.

  When setting the target throughput, there is interference between flows in actual network accommodation, so it is often not appropriate to use the maximum achievable throughput as determined above as the target value. There is also. Therefore, a value obtained by multiplying the maximum value of the throughput by α is set as the target throughput. α takes an arbitrary value in the range of 0 to 1, and α may change according to the file size. Note that α is an arbitrary value, and α = 1 in the embodiments described later. The appropriate value of α is out of the scope of the present invention because it varies depending on the type of application and the concept of quality provided to the user.

  The necessary bandwidth calculation unit 20 uses, as a necessary bandwidth, a bandwidth in which the cumulative probability is equal to or less than a predetermined value in the probability distribution of the simultaneous connection bandwidth required when all flows connected to the network maintain the target throughput. calculate.

  The necessary bandwidth calculation unit 20 includes a flow bandwidth calculation unit 21, a simultaneous connection bandwidth calculation unit 22, a total band value storage unit 23, a cumulative probability distribution storage unit 24, a cumulative probability distribution calculation unit 25, and a necessary bandwidth calculation unit 26.

  The flow bandwidth calculation unit 21 determines a probability distribution and a file size distribution that are followed by the occurrence of each flow accommodated in the network, and the target throughput of each flow determined by the throughput calculation unit 10 from the generation to the end of the flow is determined. Assuming that it is necessary, the bandwidth used by each flow is calculated.

  The simultaneous connection bandwidth calculation unit 22 calculates a total value of bandwidths used by each flow as a bandwidth used when the flows are simultaneously connected, and stores the total value in the bandwidth total value storage unit 23. In the present embodiment, as an example, it is assumed that the probability distribution followed by the occurrence of each flow follows a Poisson distribution, and the file size distribution follows an exponential distribution. However, the present invention is not limited to this. Usually, flow generation and file size distribution differ in distribution shape depending on the applied network. A heavy tail distribution is often used as a file size distribution, but this time, an exponential distribution is applied to simplify the explanation. For each flow, the simultaneous connection bandwidth is calculated on the assumption that the target throughput determined by the throughput calculation unit 10 is used from the generation to the end of the flow.

  As a method for calculating the simultaneous connection bandwidth, a time-series simultaneous use bandwidth may be calculated by numerical calculation or the like, or may be derived from a system number distribution calculated from a queue model or the like.

  The cumulative probability distribution 25 acquires the total bandwidth value used at the time of simultaneous connection of each flow obtained by the simultaneous connection bandwidth calculation unit 22 from the bandwidth total value storage unit 23, and based on the total bandwidth value, The cumulative probability distribution of the band to be used is calculated and stored in the cumulative probability distribution storage unit 24.

  The necessary band calculation unit 26 calculates a band corresponding to a predetermined arbitrary probability point N% as a necessary band for the cumulative probability distribution of the bands used in all flows acquired from the cumulative probability distribution storage unit 24. Note that the value of the probability point N% that specifies the required bandwidth on the cumulative probability distribution can be set arbitrarily. In general, 99% or 99.9% points are generally used in traffic design. The appropriate value in the present invention is outside the scope of the invention, and thus the details are not particularly limited.

[Second Embodiment]
In actual network accommodation, there is interference between flows, so it is often not appropriate to set the maximum achievable throughput value obtained in the first embodiment as the target value as it is. is there. Therefore, in the present embodiment, a value obtained by multiplying the maximum value of the throughput by α is set as the target throughput. α takes an arbitrary value in the range of 0 to 1, and α may change according to the file size. Note that α is an arbitrary value, and α = 1 in the embodiments described later. The appropriate value of α is out of the scope of the present invention because it varies depending on the type of application and the concept of quality provided to the user.

  FIG. 2 shows the configuration of the bandwidth calculation apparatus according to the second embodiment of the present invention.

  In the figure, the throughput calculation unit 30 has a throughput upper limit calculation unit 31 and a target throughput setting unit 32, which is different from the first embodiment.

The throughput upper limit calculation unit 31 calculates the maximum throughput that can be realized in a state where there is no interference between communications as the target throughput of each flow.
-Packet size (pkt);
The narrowest band on the path (BW _a );
-Buffer amount (bf _a ) in the link;
-Round-trip delay time (RTT) excluding queuing delays;
Is used to model the transition state of the TCP window size,
-Maximum congestion window size (cwnd _ss ) in the slow start phase;
-Number of retransmitted packets (R _ss ) after reaching maximum congestion window size in slow start phase;
-Maximum congestion window size (cwnd _ca ) in the congestion avoidance phase;
-Number of retransmitted packets after reaching the maximum congestion window size in the congestion avoidance phase (R _ca );
To estimate the maximum value of the throughput.

  As a method for estimating the maximum throughput, in addition to the above, by using ns-2 as a packet simulator and performing a packet simulation that simulates the behavior of TCP when an arbitrary file size is given, the flow size It is possible to easily determine the maximum feasible throughput corresponding to (for example, see “ns-2: http://nsnam.isi.edu/nsnam/index.php/Main_Page”).

  The target throughput setting unit 32 multiplies the maximum value of the throughput estimated by the throughput upper limit calculation unit 31 by α. Note that α is an arbitrary value, and an appropriate value is the same as the concept of target throughput setting in the first embodiment described above.

  The above method will be described in detail.

  FIG. 3 is a flowchart of throughput calculation processing according to the second embodiment of the present invention. In the following, steps 101 to 106 are a throughput upper limit calculation method (throughput upper limit calculation unit 31), and step 107 is a description of a function (target throughput setting unit 32) that multiplies the upper limit value of the throughput by α.

  Step 101) Modeling of TCP Window Size Transition State In this specification, for simplicity, the congestion window size is described as a value normalized by the packet size. The following is an example obtained by the method of calculating the throughput upper limit of TCP NewReno.

  TCP NewReno generally introduces DelayedACK as an option on the receiving side, and considers an upper limit of throughput when the option is valid. At this time, since the transmission side receives one ACK every time two packets are transmitted, the congestion window size increases by about 1.5 times in the slow start phase. In the congestion avoidance phase, the window size increases by one packet each time 1 RTT elapses. In the following, it is assumed that all ACK packets arrive at the same time, and transmission packets are transmitted in bursts. The burst interval is divided into time slots. Although the burst transmission interval is in principle equivalent to RTT, the length of the time slot is corrected in consideration of the queuing delay according to the burst size.

Step 102) Calculation of cwnd _ss The congestion window size in the slow start phase follows the operation of the ACK, and the congestion window size cwnd _i in the i-th time slot is expressed as follows.

なお、初期値としてcwnd₁=2，n₁=2，m₁=1を与える。

Note that cwnd ₁ = 2, n ₁ = 2 and m ₁ = 1 are given as initial values.

Here, the value of cwnd _ss is obtained as follows according to the parameter input in step 101.

・ (Condition 1) When RTT * BW _a ≤ 2 * bf _a

となる。

It becomes.

・ (Condition 2) When RTT * BW _a > 2 * bf _a

となる。

It becomes.

Step 103) Calculation of R _ss
Since R _ss becomes a value depending on cwnd _ss , it becomes a value according to the input parameter in step 101 also in this step.

(Condition 1) When RTT * BW _a ≤ 2 * bf _{a According} to the derivation of cwnd _ss in step 102, the number of packet retransmissions R _ss that occurs at the transition from the slow start phase to the congestion avoidance phase is

となる。

It becomes.

(Condition 2) When RTT * BW _a > 2 * bf _{a According} to the derivation of cwnd _ss in step 102, the number of packet retransmissions R _ss that occurs at the transition from the slow start phase to the congestion avoidance phase is

となる。

It becomes.

Step 104) Calculation of the threshold value ssthresh _ss of the slow start phase and the congestion avoidance phase:
Ssthresh _ss , which is the threshold for the slow start phase and congestion avoidance phase, is

となる。

It becomes.

Step 105) Calculation of cwnd _ca and R _ca :
In the congestion avoidance phase, two packets are sent in principle every time one ACK is received, and one more packet is sent when the congestion window size increases by one packet. In other words, the congestion window size increases by one packet each time one time slot elapses. Also, the congestion window size in the i-th slot, the number of transmitted packets and the number of ACK packets transmitted on the receiving side are:

と表される。

It is expressed.

  In the congestion avoidance phase, the congestion window size is

になるまで大きくなる。また、パケットの再送数R_caは1パケットとなる。

Grows until Also, the packet retransmission number R _ca is 1 packet.

The threshold ssthresh _ca with the next congestion avoidance phase is

となる。

It becomes.

Step 106) Estimating Throughput The total value of the number of transmitted packets in each slot calculated in Steps 102 to 105 is compared with the number of transmitted packets estimated from the file size of the flow, and the time required to complete the flow is obtained. . The throughput is calculated by dividing the flow size by the time.

Step 107) Target throughput setting:
In actual network accommodation, since there is interference between flows, it is often not appropriate to set the maximum achievable throughput value obtained above as a target value as it is. Therefore, a value obtained by multiplying the maximum value of the throughput by α is set as the target throughput. α takes an arbitrary value in the range of 0 to 1, and α may change according to the file size.

  The above embodiment will be specifically described below.

[First embodiment]
First, the throughput calculator 10 in the first embodiment will be described.

  In this embodiment, the target throughput per flow is uniformly set to 10 Mbps, but these values are set in consideration of applications provided by TCP, flow characteristics, and the like. For example, in applications that transmit video over TCP, the video encoding rate or the like may be the target throughput, and in file downloads, the target throughput may be within a certain period of time. . Moreover, the target value may differ for every flow.

  Next, the processing of the bandwidth calculation unit 20 will be described.

  FIG. 4 is a flowchart of the operation of the bandwidth calculation unit of the first exemplary embodiment of the present invention.

Step 201) Calculate the simultaneous use bandwidth of the flow:
The flow band calculation unit 21 calculates a required band when it is assumed that the file size distribution follows an exponential distribution with an average of 5 MB, and 1000 flows occur randomly in 100 seconds. FIG. 5 shows an example of a band required in time series.

Step 202) Create cumulative probability distribution of used bandwidth:
A cumulative probability distribution of the used band is calculated from the simultaneous used band obtained in step 201. FIG. 6 shows a complementary distribution of the cumulative probability distribution corresponding to FIG.

Step 203) Required bandwidth calculation:
The required bandwidth calculation unit 26 calculates the required bandwidth for the cumulative probability distribution of the used bandwidth obtained in step 202. According to FIG. 6, when 99.9% is adopted as the N% point, the necessary bandwidth is 590 Mbps.

[Second Embodiment]
A specific description will be given along the flowchart of FIG. 3 of the second embodiment described above.

  First, the processing of the throughput calculation unit 30 will be described.

Step 101) Input of parameters:
The throughput upper limit calculation unit 31 of the throughput calculation unit 30 has a packet size pkt of 1500 bytes, the narrowest bandwidth BW _a on the path is 100 Mbps, the buffer amount bf _a on the link is 50 packets, and a round trip delay time excluding the queuing delay RTT Is set to 10 ms, and the achievable flow throughput under these conditions is calculated below.

Step 102) Calculation of cwnd _ss :
With this step, the congestion window size in the slow start phase of TCP is

と表される。また、本実施例では、ステップ１０１のパラメータにより、

It is expressed. In this embodiment, according to the parameter of step 101,

となることから、（条件１）RTT*BW_a <= 2*bf_a が成立するため、スループット上限算出部３１は、第２の実施の形態で示した式（1），（2）より、i=12，cwnd_ss=203を求める。

(Condition 1) Since RTT * BW _a ≤ 2 * bf _a is established, the throughput upper limit calculation unit 31 obtains the following expressions (1) and (2) from the second embodiment: Find i = 12, cwnd _ss = 203.

Step 103) Calculation of R _ss :
As described in Step 102, (Condition 1) Since RTT * BW _a <= 2 * bf _a , R _ss = 70 from Equation (4).

Step 104) Calculation of the threshold value ssthresh _ss of the slow start phase and the congestion avoidance phase:
Since cwnd _ss is given in step 102, ssthresh _ss is 101.5 from equation (6).

Step 105) Calculation of cwnd _ca and R _ca :
In this step, the maximum congestion window size in the congestion avoidance phase and the number of retransmission packets after reaching the maximum value are obtained. In this embodiment, cwnd _ca = 133 from equation (8), and R _ca = 1 from equation (10).

Step 106) Throughput estimation:
From the above steps 101 to 105, the elapsed time necessary for transferring the file size of each flow is calculated, and the throughput is estimated. The result of the estimated throughput is shown in FIG.

Step 107) Target throughput setting:
The target throughput setting unit 32 adopts 1 as α and sets the value calculated in step 106 as the target value.

  Next, the processing of the necessary bandwidth calculation unit 20 will be described along the second embodiment.

  This will be described with reference to FIG.

Step 201) Calculate the simultaneous connection bandwidth of the flow:
The simultaneous connection bandwidth calculation unit 22 calculates a required bandwidth when it is assumed that a file size distribution follows an exponential distribution with an average of 5 MB, and 1000 flows occur randomly in 100 seconds. FIG. 8 shows an example of a time-series change of a necessary band.

Step 202) Create cumulative probability distribution of simultaneous connection bands:
The cumulative probability distribution calculation unit 25 calculates the cumulative probability distribution of the used band from the simultaneous used band obtained in step 201. FIG. 9 shows a complementary distribution of the cumulative probability distribution corresponding to FIG.

Step 203) Required bandwidth calculation:
The required bandwidth calculation unit 26 calculates the required bandwidth for the cumulative probability distribution of the used bandwidth obtained in step 202. From FIG. 9, if 99.9% is adopted as the N% point, the required bandwidth is 904 Mbps.

<Evaluation of estimation accuracy>
The estimation accuracy of the flow throughput that can be executed for each flow size was evaluated. Throughput in a network accommodating only a single flow was evaluated by simulation. FIG. 10 shows the estimated flow throughput obtained by the present invention. As shown in FIG. 10, the flow throughput obtained by the present invention is in good agreement with the simulation result, and it can be confirmed that the estimation method is highly accurate.

  In addition, the operation | movement of each component of the zone | band calculation apparatus shown in said 1st, 2nd embodiment is constructed | assembled as a program, installed in the computer utilized as a zone | band calculation apparatus, and is made to perform, or a network is used. It is possible to circulate through.

  The present invention is not limited to the above-described embodiments and examples, and various modifications and applications are possible within the scope of the claims.

DESCRIPTION OF SYMBOLS 10 Throughput calculation part 20 Necessary band calculation part 21 Flow band calculation part 22 Simultaneous connection band calculation part 23 Band total value storage part 24 Cumulative probability distribution storage part 25 Cumulative probability distribution calculation part 26 Necessary band calculation part 30 Throughput calculation part 31 Throughput upper limit Calculation unit 32 Target throughput setting unit

Claims (7)

A receiving terminal having a function of returning a reception confirmation packet to the transmission terminal when the reception terminal receives a data packet from the transmission terminal via a network that performs communication by packet exchange, and the reception terminal receives the reception confirmation packet. A bandwidth calculating device in a network that includes a communication terminal that has a function of grasping the amount of data, and that includes a communication system in which a window size increases every time a reception confirmation packet from the reception terminal arrives,
Throughput calculation means for setting a target throughput for each flow characteristic for each flow accommodated in the network;
Bandwidth calculating means for calculating a bandwidth in which the cumulative probability of the target throughput total value of each flow accommodated in the network is a certain value or less;
A bandwidth calculation apparatus comprising: The throughput calculating means includes
The band calculation device according to claim 1, further comprising means for giving α times the maximum throughput achievable in a state where there is no interference between communications as a target throughput for each flow. The throughput calculating means includes
When calculating the maximum throughput that can be achieved without interference between communications, using the packet size, the narrowest bandwidth on the path, the buffer amount in the link, and the round trip delay time excluding the queuing delay, Find the maximum congestion window size in the slow start phase, the number of retransmission packets after reaching the maximum congestion window size in the slow start phase, the maximum congestion window size in the congestion avoidance phase, the number of retransmission packets after reaching the maximum congestion window size in the congestion avoidance phase, The bandwidth calculating apparatus according to claim 2, further comprising means for estimating a maximum value of the throughput. A receiving terminal having a function of returning a reception confirmation packet to the transmission terminal when the reception terminal receives a data packet from the transmission terminal via a network that performs communication by packet exchange, and the reception terminal receives the reception confirmation packet. A bandwidth calculation method in a network including a communication system having a transmission terminal having a function of grasping a data amount, and a window size increasing each time a reception confirmation packet from the reception terminal arrives,
Throughput calculating means for setting a target throughput for each flow characteristic for each flow accommodated in the network;
A bandwidth calculating step in which the bandwidth calculating means calculates a bandwidth in which the cumulative probability of the target throughput total value of each flow accommodated in the network is a certain value or less;
A bandwidth calculation method characterized by: In the throughput calculation step,
The bandwidth calculation method according to claim 4, wherein α times the maximum throughput that can be realized in a state where there is no interference between communications is given as a target throughput for each flow. In the throughput calculation step,
When calculating the maximum throughput that can be achieved without interference between communications, using the packet size, the narrowest bandwidth on the path, the buffer amount in the link, and the round trip delay time excluding the queuing delay, Find the maximum congestion window size in the slow start phase, the number of retransmission packets after reaching the maximum congestion window size in the slow start phase, the maximum congestion window size in the congestion avoidance phase, the number of retransmission packets after reaching the maximum congestion window size in the congestion avoidance phase, The bandwidth calculation method according to claim 5, wherein the maximum value of the throughput is estimated.   The bandwidth calculation program for functioning a computer as each means which comprises the bandwidth calculation apparatus of any one of Claim 1 thru | or 3.

Images