JP2005057620A - Transmission quality control method and its device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a transmission quality control method achieving a packet discarding method for maintaining TCP transmission quality while suppressing a queuing delay in a buffer which affects UDP transmission quality, and its device. <P>SOLUTION: The speed of a relay link is set to C, and a maximum window size in TCP transmission is set to Wmax. A TCP user i's access link speed Ca(i) multiplexed on the relay link, and a minimum value RTTmin(i) of a reciprocating propagation delay time in the TCP transmission of the user i, are held beforehand. When a packet from the user i arrives at a buffer, the packet is discarded if an output waiting queue length Q (byte) in the buffer is over the value of (Wmax/Ca(i)-RTTmin(i)) * C. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a communication quality control method and apparatus effective for maintaining and managing TCP communication quality and UDP communication quality by efficiently using a buffer in an IP network.

  As IP networks are widely used, there is an increasing demand for communication quality assurance on IP networks. The most common buffer management in the current IP network is Drop-Tail, in which packets are accumulated until the prepared buffer is full, and when the buffer is full, the packet that arrives at that time is discarded.

  In this method, packets are accumulated until the buffer is full. When the network is congested, the queuing delay and fluctuation in the buffer become very large, and both TCP communication and UDP communication that transfers stream / voice applications are performed. Will degrade the performance. In particular, since the real-time property is important for UDP communication quality, it is greatly affected by the delay time.

  As a packet discarding method for improving the fairness between the TCP throughput and the TCP flow, RED (Randam Early Detection) (for example, see Non-Patent Document 1) is generally known. RED is based on a congestion detection method that uses an average queue length to prevent bursty packet discard by performing stochastic packet discard. This method has a number of parameters to be set, and depending on the set value, there is a problem in that the performance may be worse than Drop Tail depending on the traffic condition (for example, Non-Patent Document 2). reference).

  In addition, several buffer management methods for improving the fairness between the TCP throughput and the TCP flow have been proposed (see, for example, Patent Documents 1 and 2). However, both of these methods are aimed at improving only the TCP quality, and when the UDP traffic is mixed, the influence on the UDP communication quality is considered and the queuing delay in the buffer is suppressed. There was a problem that this was not considered.

  On the other hand, as a conventional buffer design method, there has been proposed a method for calculating a buffer size necessary for suppressing a packet (or ATM communication cell) loss rate to a predetermined target value or less (for example, Patent Document 3). reference). However, the quality target value in this method is at the IP level (or ATM level), and is not a design method that directly considers the TCP quality.

JP-T-2001-522183 JP-A-10-233802 JP-A-3-235550 S. Floyd and V. Jacobson, "Random early detection gateways for congestion avoidance," IEEE / ACM Transaction on Networking, vol.1, pp. 397-413, Aug. 1993. Hasegawa, Itaya, Murata “Dynamic threshold control method of RED in backbone router”, IEICE Technical Report (NS2001-11), April 2001

  In view of the above-described problems, an object of the present invention is to provide a communication quality control method capable of realizing a packet discarding method that can maintain TCP communication quality while suppressing queuing delay in a buffer that affects UDP communication quality. And providing an apparatus thereof. Another object of the present invention is to provide a buffer design method capable of reducing useless buffer resources by calculating a buffer size necessary for maintaining TCP communication quality using only parameters that can be easily set. Is to provide a communication quality control method and apparatus capable of realizing the above.

  In order to achieve the above object, a communication quality control method and apparatus according to the present invention are characterized by utilizing the TCP traffic characteristics described below.

According to the study of the present inventor, the following TCP traffic characteristics have been found.
That is, assuming that a TCP connection is multiplexed on a relay link having a link speed of C, the maximum window size in TCP communication is Wmax, the access link speed of a TCP user multiplexed on the relay link is Ca, and the round-trip propagation delay When the minimum time value is RTTmin, when the buffer size Bf (or TCP packet discard threshold) is larger than (Wmax / Ca-RTTmin) * C [bytes], the file transfer time in the TCP connection passing through the relay link The link utilization rate at which the average value of the link starts to deteriorate is constant regardless of the value of Bf.

  FIG. 1 shows an example of the evaluation. This is C = 22.5 Mbps × {512 bytes / (512 + 40) bytes} (TCP data segment size = 512 bytes, TCP / IP header size = 40 bytes, and the link speed 22.5 Mbps is corrected in consideration of the overhead), Ca = 2 Mbps × {512 bytes / (512 + 40) bytes} This is a result of a simulation experiment in which a TCP connection generated according to the Poisson arrival process downloads a file according to an exponential distribution with an average of 10 KB.

Further, Wmax = 8 KB and RTTmin = 2.68 ms were set. Here, the case of Bf = 162 packets (the packet size is 512 bytes, so Bf = 162 × 512 bytes in terms of bytes) corresponds to (Wmax / Ca−RTTmin) * C [byte].
From this, it can be seen that when Bf> 162 packets, the deterioration point with respect to the link usage rate of the file transfer time is constant regardless of Bf, and conversely, when Bf <162 packets, the smaller the Bf, the lower the deterioration point.

  This phenomenon depends on the load until the average file transfer time exceeds a certain link usage rate when the access line speed Ca of each user multiplexed on a certain relay link is sufficiently smaller than the link speed C. There are two characteristics: a characteristic that the buffer size is constant, and a deterioration point that is constant when the buffer size is larger than (Wmax / Ca-RTTmin) * C [byte].

This characteristic can be explained as follows. In other words, the quality of TCP depends on two factors: delay and packet loss. First, consider a case where there is no packet loss. In a TCP connection having a sufficiently large file size so that the influence of TCP slow start can be ignored, the TCP throughput (Th) is given as follows.
Th = min (Ca, Wmax / RTT)
RTT = Dq + RTTmin
Here, Dq is a queuing delay.

Thus, when Dq> (Wmax / Ca−RTTmin), TCP cannot achieve the ideal throughput Ca, and the quality deteriorates. Therefore, when Dq increases, TCP quality deteriorates even if packet loss does not occur. That is, if a buffer large enough to satisfy Dq> (Wmax / Ca−RTTmin) is prepared, the quality degradation point is constant regardless of the buffer size (see FIG. 1).
On the other hand, if the buffer size is smaller than (Wmax / Ca−RTTmin) * C [byte], quality degradation due to packet loss occurs before TCP quality degradation due to delay starts. Therefore, the smaller the buffer size, the smaller the quality degradation point. (See FIG. 1).

  Further, when the average queuing delay time is compared between Bf = ∞ and Bf = 162 [packets], the delay time diverges when Bf = ∞, whereas Bf = 162 [packets]. Can be suppressed to about 20 ms (<Wmax / Ca-RTTmin) (see FIG. 2). Therefore, the queue length can be suppressed to (Wmax / Ca-RTTmin). Therefore, if the queue length is suppressed to (Wmax / Ca−RTTmin) * C [byte] or less, it is possible to suppress the queuing delay that affects the UDP communication without affecting the TCP quality.

  Using the above TCP characteristics, the first communication quality control method according to the present invention assumes that a TCP connection and a UDP connection are multiplexed on a certain relay link in the communication network. A communication quality control method for discarding packets according to a queue length, wherein the speed of the relay link is C, the maximum window size in TCP communication is Wmax, and the access link speed of a TCP user i multiplexed on the relay link Ca (i) and the minimum value RTTmin (i) of the round trip propagation delay time in the TCP communication of the user i are stored in advance, and when a packet from the user i arrives at the buffer, an output waiting queue in the buffer If the length Q [byte] exceeds (Wmax / Ca (i) −RTTmin (i)) * C, the packet is discarded. It is characterized in.

  Here, the value of RTT (i) takes various values for each user and for each communication. However, if the value is set to a smaller value (or 0), the threshold value becomes larger, and the threshold value setting on the safe side is set. It becomes.

  The second communication quality control method according to the present invention assumes that a TCP connection and a UDP connection are multiplexed on a certain relay link in a communication network, and discards a packet according to the queue length of an output waiting packet buffer of the relay link. A communication quality control method, wherein the speed of the relay link is C, the maximum window size in TCP communication is Wmax, the minimum value Ca_min of the user access link speed multiplexed on the relay link, and the TCP communication of the user i The minimum value RTTmin_min among the minimum values RTTmin (i) of the round-trip propagation delay time in FIG. 6 is grasped, and when a packet from the TCP user arrives at the buffer, the output waiting queue length Q [byte] in the buffer is (Wmax / Ca_min−RTTmin_min) * If it exceeds C, discard the packet. It is characterized by.

  In the third communication quality control method according to the present invention, the TCP packet is discarded by either the first or second method, and the allowable delay time Dmax_UDP in the buffer for UDP communication is determined in advance. When the UDP packet arrives at the buffer, if the output waiting queue length Q [byte] in the buffer exceeds Dmax_UDP * C, the packet is discarded.

  The fourth communication quality control method according to the present invention assumes that only a TCP connection is multiplexed on a certain relay link in the communication network, and when a packet arrives at the output waiting packet buffer of the relay link, A communication quality control method for discarding a packet when there is no free space, wherein the speed of the relay link is C, the maximum window size in TCP communication is Wmax, and the user access link multiplexed on the relay link The smallest value RTTmin_min among the minimum speed Ca_min and the minimum round-trip propagation delay time in each user's TCP communication is grasped, and the buffer size is set to (Wmax / Ca_min−RTTmin_min) * C It is characterized by designing.

  On the other hand, a communication quality control device according to the present invention is a device for performing communication quality control, and means for determining a packet discard threshold by the method according to any one of the first to third aspects, and outputs when a packet arrives A means for measuring a waiting queue length Q [bytes] and a means for discarding a packet by the method according to any one of claims 1 to 3 are provided.

  According to the present invention, a packet discarding method that can maintain TCP communication quality while suppressing queuing delay in a buffer that affects UDP communication quality, or a buffer size required to maintain TCP communication quality. Therefore, it is possible to provide an effective communication quality control method and apparatus based on a buffer design method capable of realizing useless buffer resource reduction by calculating using only parameters that are easy to set.

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

  FIG. 3 is a diagram showing an example of a basic configuration of an IP network to which the present invention is applied. The communication quality control apparatus according to the present embodiment is used in the form of being deployed in the ingress node in FIG. 4 is a block diagram showing a configuration example of a communication quality control apparatus using the communication quality control method of the present invention, and FIGS. 5 to 8 are operation flowcharts for explaining the operation of each unit shown in FIG. is there.

  As shown in FIG. 4, the communication quality control apparatus according to this embodiment includes a user information management unit 11, a queue length monitoring unit 12, a packet discard determination unit 13, a discard threshold calculation unit 14, and a packet storage unit (buffer). 15 and a packet transfer unit 16.

  A packet arriving from the user is transmitted to the user information management unit 11, the queue length monitoring unit 12, and the packet discard determination unit 13. When the packet arrives in the procedure shown in FIG. 5 (step 21), the user information management unit 11 analyzes the packet header information (step 22), and analyzes the header part to determine which user the packet is from. If it is determined that the packet is from the user i (step 23), the access speed Ca (i) and the minimum round-trip propagation delay time value RTTmin (i) are stored as information on the user i from the management table 11a prepared in advance. ) And notify the discard threshold value calculation unit 14.

  FIG. 9 illustrates a part of the contents of the management table 11 a prepared in the user information management unit 11. As shown in FIG. 9, in the management table 11a, the access speed (Ca (i)) of the user, the minimum value of the round trip propagation delay time (RTTmin (i)), etc. are stored in advance for each user. Has been.

  The discard threshold value calculation unit 14 performs the procedure as shown in FIG. 6 (step 31) Ca (i), RTTmin (i) notified from the user information management unit 11, and the maximum window size Wmax stored in advance. Then, using the relay link speed C (step 32), in step 33, the packet discard threshold is calculated by (Wmax / Ca (i) -RTT (i)) * C, and this threshold is notified to the packet discard determination unit 13 (Step 34).

  On the other hand, the queue length monitoring unit 12 that has received the packet from the user reads the packet size for the arrived packet (in the case of “arrival” in step 41) by the procedure shown in FIG. [Byte] is updated by Q ← Q + {the packet size} (step 42), and the new Q value is notified to the packet discard determination unit 13 (step 43).

  When the packet discard determination unit 13 receives the packet (step 51), the Q value notified from the queue length monitoring unit 12 and the threshold value notified from the discard threshold calculation unit 14 (step 51) Wmax / Ca (i) -RTT (i)) * C values are compared (steps 52 and 53), Q> (Wmax / Ca (i) -RTT (i)) * C, and the user packet is TCP If so, the received packet is discarded (step 54).

  Further, the result of comparison between the above-described Q value and the threshold value (Wmax / Ca (i) −RTT (i)) * C is Q <(Wmax / Ca (i) −RTT (i)) * C, Alternatively, if the user packet is not TCP, the user packet is accepted and sent to the packet storage unit (buffer) 15. The packet transfer unit 16 takes out the packet from the packet storage unit 15 and performs a transfer process (step 55).

  Thereafter, the packet transfer unit 16 checks whether or not there is still a packet in the packet storage unit 15. If there is, the packet transfer unit 16 takes out the next packet from the packet storage unit 15 and performs transfer processing. Repeat until there are no more packets in section 15. Further, when the packet transfer unit 16 reads a packet, the packet transfer unit 16 notifies the queue length monitoring unit 12 of the size of the packet. When receiving the packet size from the packet transfer unit 16, the queue length monitoring unit 12 updates the queue length Q by Q ← Q− {the packet size}.

  The communication quality control method and apparatus of the present invention have been described in detail above. However, the present invention is not limited to the above-described embodiment, and various modifications and improvements can be made without departing from the spirit of the present invention. It goes without saying that you can go.

  Further, the communication quality control method of the present invention can be used effectively when calculating the buffer size necessary for maintaining the TCP communication quality using only parameters that can be easily set, This also provides a new buffer design method capable of reducing useless buffer resources.

It is a figure explaining the TCP characteristic based on this invention, and is a graph which shows the result of having measured the average file transfer time vs link utilization rate by the simulation experiment when changing the buffer size. FIG. 6 is a graph showing the results of actual measurement of average queuing delay time vs link usage rate through simulation experiments. FIG. It is a figure which shows the basic structural example of the network to which this invention is applied. It is a block diagram which shows the structural example of the communication quality control apparatus in this invention. FIG. 5 is an operation flowchart for explaining the operation of the user information management unit 11 of the communication quality control apparatus shown in FIG. 4. FIG. 5 is an operation flowchart for explaining the operation of a discard threshold value calculation unit 14 of the communication quality control device shown in FIG. 4. FIG. 5 is an operation flowchart for explaining the operation of the queue length monitoring unit 12 of the communication quality control device shown in FIG. 4. FIG. 5 is an operation flowchart for explaining the operation of the packet discard determination unit 13 of the communication quality control device shown in FIG. 4. It is a figure which shows the structural example of the management table 11a provided in the user information management part 11 of the communication quality control apparatus shown in FIG.

Explanation of symbols

DESCRIPTION OF SYMBOLS 11 User information management part 12 Queue length monitoring part 13 Packet discard determination part 14 Discarding threshold calculation part 15 Packet accumulation part (buffer)
16 packet transfer part 21-25, 31-34, 41-44, 51-55 Processing step

Claims (5)

A communication quality control method in which a TCP connection and a UDP connection are multiplexed on a certain relay link in a communication network, and a packet is discarded according to a queue length of an output waiting packet buffer of the relay link,
The speed of the relay link is C, the maximum window size in TCP communication is Wmax, the access link speed Ca (i) of the TCP user i multiplexed on the relay link and the round-trip propagation delay time in the TCP communication of the user i The minimum value RTTmin (i) is held in advance,
When the packet from the user i arrives at the buffer, if the output waiting queue length Q [byte] in the buffer exceeds (Wmax / Ca (i) −RTTmin (i)) * C, the packet A communication quality control method characterized by discarding the communication. A communication quality control method in which a TCP connection and a UDP connection are multiplexed on a certain relay link in a communication network, and a packet is discarded according to a queue length of an output waiting packet buffer of the relay link,
The speed of the relay link is C, the maximum window size in TCP communication is Wmax, the minimum value Ca_min of the user's access link speed multiplexed on the relay link and the minimum value of the round-trip propagation delay time in the TCP communication of the user i Know the smallest value RTTmin_min in RTTmin (i),
When a packet from a TCP user arrives at a buffer, if the output waiting queue length Q [byte] in the buffer exceeds (Wmax / Ca_min−RTTmin_min) * C, the packet is discarded. Communication quality control method. Discarding the TCP packet by the method according to claim 1 or 2;
In addition, an allowable delay time Dmax_UDP in the buffer for UDP communication is determined in advance,
A communication quality control method characterized in that, when a UDP packet arrives at the buffer, if the output waiting queue length Q [byte] in the buffer exceeds Dmax_UDP * C, the packet is discarded. Assume that only a TCP connection is multiplexed on a certain relay link in the communication network,
When a packet arrives at the output waiting packet buffer of the relay link, if there is no space in the buffer, a communication quality control method for discarding the packet,
The speed of the relay link is C, the maximum window size in TCP communication is Wmax,
Knowing the minimum value Ca_min of the user access link speed multiplexed on the relay link and the smallest value RTTmin_min among the minimum values of the round-trip propagation delay time in the TCP communication of each user,
A communication quality control method, wherein the buffer size is designed by setting the buffer size to (Wmax / Ca_min−RTTmin_min) * C. An apparatus for performing communication quality control,
Means for determining a packet discard threshold by the method according to any one of claims 1 to 3;
Means for measuring the output waiting queue length Q [bytes] when a packet arrives;
A communication quality control apparatus comprising means for discarding a packet by the method according to any one of claims 1 to 3.

Images