JP2005117131A - TCP traffic control method and control apparatus - Google Patents

Abstract

[PROBLEMS] To effectively use network resources even during network congestion, without requiring special signaling protocol or state management for each connection, and avoiding continuation of congestion, promotion, and invalid suspension of resources. TCP quality is ensured.
When a TCP connection is multiplexed on a link between nodes in a communication network, when a packet arrives at a header analysis unit 41 of the node, the start of the TCP connection is indicated in the TCP header of the packet. When the SYN flag is set and the traffic monitoring unit 44 detects that the link is congested, the packet is discarded by notifying the packet discard determining unit 42. As another method, the link output waiting buffer arranged in the packet transfer unit 43 is divided into two classes, the SYN flag indicating the start of the TCP connection is set in the TCP header of the packet, and the link is congested. If so, it is accommodated in the low priority class buffer, otherwise it is accommodated in the high priority buffer.
[Selection] Figure 2

Description

  The present invention relates to a TCP traffic control method and control apparatus capable of maintaining TCP communication quality during communication even when the network is congested by effectively using network resources in an IP network.

As IP networks are widely used, there is an increasing demand for communication quality assurance over IP networks. Many Internet-related protocols are the only type of protocol that does its best to establish communication, and this is the reason why this policy can realize the network at a low cost. Therefore, in an IP network, the best effort type communication that does not reserve a network resource at the start of communication is common.
In this case, all TCP (Transmission Control Protocol) connections for newly starting communication are accepted in the network. When the network is congested, the TCP quality is significantly degraded in inverse proportion to the number of simultaneous connection connections in the bottleneck link / node. Eventually, the quality of all users will deteriorate.

  Further, since packet loss and delay variation occur during network congestion, there is a possibility that the number of retransmitted packets due to TCP increases, and wasteful traffic increases, thereby promoting congestion. Furthermore, when the TCP quality is significantly deteriorated, some users give up communication on the way, and packet transfer that has been spent for that user becomes useless until that, and there is a possibility that invalidation of network resources may occur.

  The above-mentioned problems are as follows. Massouli and J.M. Roberts, “Arguments in favor of admission control for TCP FLOWS,” ITC 16, 1999. Although it is pointed out also in (refer nonpatent literature 1), the concrete solution method is not shown. Non-Patent Document 1 describes a method for managing the state for each TCP connection and limiting the number of received connections. However, a node with a large number of multiple connections has a problem in scalability.

  On the other hand, there is a protocol called RSVP (Resource Reservation Protocol) that performs bandwidth reservation for each router in the network using bandwidth reservation signaling at the start of communication. This is a protocol for managing the transmission band of a specific communication channel between routers, and is a method of securing the band of the WAN side line part that is likely to become a bottleneck. However, this method requires each router in the network to perform RSVP signaling processing, and each router needs to manage the state for each TCP connection. There was a problem with scalability.

JP-T-2001-522183 JP-A-10-233802 JP 2000-49853 A Japanese Patent Laid-Open No. 2003-23443 JP 2001-127830 A L. Massouli and J.M. Roberts, “Arguments in favor of admission control for TCP FLOWS,” ITC 16, 1999. S. Floyd and V.M. Jacobson, “Random early detection gateways for consonance aviation,” IEEE / ACM Transaction on Networking, vol. 1, pp. 397-413, Aug. 1993. Kawahara, Ishibashi, Asaka, Ori, “IP Traffic Management Method Using TCP Flow Statistics”, IEICE Technical Report NS2002-170, pp. 53-56, 2002.

Many proposals have been made regarding TCP traffic control. RED (Random Early Detection) is generally known as a packet discarding method for improving the fairness between the TCP throughput and the TCP connection. RED is, for example, S.I. Floyd and V.M. Jacobson, “Random early detection gateways for consonance aviation,” IEEE / ACM Transaction on Networking, vol. 1, pp. 397-413, Aug. 1993 (see Non-Patent Document 2).
RED is based on a congestion detection method that uses an average queue length to prevent bursty packet discard by performing stochastic packet discard.

In addition, several buffer management methods for improving the fairness between the TCP throughput and the TCP connection have been proposed. For example, Japanese translations of PCT publication No. 2001-522183 (refer patent document 1), Unexamined-Japanese-Patent No. 10-233802 (refer patent document 2), Unexamined-Japanese-Patent No. 2000-49853 (refer patent document 3), and Unexamined-Japanese-Patent No. 2003-23443. There is a publication (see Patent Document 4).
However, both are for ensuring fairness between TCP connections, and do not limit the number of communication request connections. Therefore, when a large number of connections occur, the TCP quality of all users deteriorates significantly. There was a problem that.

  Similar to the present invention, there is a method of performing traffic control using TCP flag information (for example, Japanese Patent Laid-Open No. 2001-127830 (see Patent Document 5)). These methods are related to a data communication method and a data communication network that perform high-performance data transfer at the time of congestion, in particular, at the time of congestion in the direction indicating delivery confirmation (ACK), etc., and the TCP header indicates ACK In this case, a desired target is realized by giving priority to the transfer. Since these methods do not limit the number of communication request connections themselves, there is a problem that when a large number of connections are generated, the TCP quality of all users is remarkably deteriorated.

(the purpose)
An object of the present invention is to solve the above-mentioned problems, and effectively use network resources even in the event of network congestion, without requiring a special signaling protocol or state management for each connection, and continuing or promoting congestion. It is another object of the present invention to provide a TCP traffic control method and control apparatus that can avoid invalid hold of resources and can secure TCP quality during communication.

The present invention is a method for limiting the number of accepted connections by discarding a packet having a SYN flag indicating the start of TCP communication when the network is congested. The SYN flag is set without managing the state for each connection. Control can be performed depending on whether or not.
In the TCP traffic control method according to the first embodiment of the present invention, when a TCP connection is multiplexed on a certain link between nodes in a communication network, when the packet arrives at the node, the TCP header of the packet If the SYN flag indicating the start of the TCP connection is set in the link, and the link is detected to be congested, the packet is discarded.

In the TCP traffic control method according to the second embodiment of the present invention, the link output waiting buffer is divided into two classes, and priority control is performed by extracting packets from each class buffer according to a predetermined weight. When a packet arrives at a node, if a SYN flag indicating the start of a TCP connection is set in the TCP header of the packet and it is detected that the link is congested, The packet is accommodated in the low priority class buffer, and if not, the packet is accommodated in the high priority buffer for priority control.
In the TCP traffic control method according to the third embodiment of the present invention, instead of performing selective discard or priority control of a packet in which the SYN flag is set by any one of the first to second embodiments. , Only the packets flagged in both SYN and ACK are subjected to selective discarding or priority control of packets by any of the methods of the first to second embodiments.
In this method, so-called client-server type communication is assumed, a SYN packet that is the start of communication is transmitted from the client to the server, and the server that has received the packet transmits a packet flagged to SYN and ACK to the client. Reply, large data is downloaded from the server to the client. Therefore, since the direction in which a packet flagged in SYN and ACK is transmitted matches the direction in which a large amount of data is transferred, the SYN and ACK packets that are the first packets that cause a large amount of data are selectively discarded. As a result, finer grain control is possible.

In the TCP traffic control method according to the fourth embodiment of the present invention, a table for managing the destination IP address of a user who has transmitted a packet flagged in SYN and ACK is prepared, and a flag is set in SYN and ACK. If the packet is received without being discarded by a method described later when a packet with a flag is received, the destination IP address #i of the packet is read, and the IP address #i is already entered in the table If so, the SYN packet passage counter C pass (i) is incremented by one, and if it is a new packet not entered in the table, it is added to the table and the SYN packet passage counter C pass (i) is set to one. To do. When a predetermined timeout time Tout has elapsed since the counter was last updated, C pass (i) ← INT [C pass (i) × α] (α is a parameter satisfying 0 <α <1, INT [ x] is a maximum integer not exceeding x), and when Cpass (i) = 0, the address #i is deleted from the table.
When a packet whose destination IP address is j arrives at a node, if the SYN and ACK in the TCP header of the packet are flagged and it is detected that the link is congested, the packet belongs to If the destination IP address #i is not entered in the table, the packet is accepted. If the destination IP address #j is already entered in the table, the counter value Cpass ( j) is read from the table and the packet is discarded with probability P inc (C pass (j)) (where P inc (x) is 0 ≦ P inc (x) ≦ 1 and monotonically increasing function with respect to x)) It is characterized by doing.

In the method of the fourth embodiment, so-called client / server communication is assumed, and a SYN packet, which is the start of communication, is transmitted from the client to the server, and the server that receives it sends a flag to SYN and ACK to the client. Returns a packet with At this time, the IP address of the client is set as the destination IP address of the packet.
Therefore, in the present invention, the access status of the client is grasped by managing the destination IP address of the packet flagged in SYN and ACK. The SYN packet that is first transmitted from the client to the server and the source IP address of the packet that represents the IP address of the client at that time may be managed, but in this case, the transmission direction of the SYN packet and the server Since the direction of so-called heavy traffic downloaded to the client is reversed, the direction in which link congestion occurs and the direction in which the SYN packet to be managed flows are reversed.
Here, it is possible to improve the fairness among users of the control method in the methods of the first to sixth embodiments by preferentially discarding users who have received many SYN packets without being discarded. Become. Although it is necessary to create a table for each connection, since only the SYN packet is targeted, it is considered that the processing load is lighter than the management for each connection for grasping the number of connections (see Non-Patent Document 1 above).

In the TCP traffic control method according to the fifth embodiment of the present invention, a table is prepared for managing the destination IP address of a user who has transmitted a packet flagged in SYN and ACK, and flags SYN and ACK. If the packet is discarded by a method to be described later when a packet having a packet is received, if the destination IP address #i of the packet has already been entered in the table, the SYN packet discard counter C drop (i ) Is incremented by 1, and if it is a new packet not entered in the table, it is added to the table and the counter C drop (i) is set to 1. When a predetermined timeout time Tout has elapsed since the counter was last updated, C drop (i) ← INT [C drop (i) × α] (α is a parameter satisfying 0 <α <1, INT [ x] is the largest integer not exceeding x), and when C drop (i) = 0, the address #i is deleted from the table.
When a packet whose destination IP address is j arrives at a node, if a flag is set in SYN and ACK in the TCP header of the packet and it is detected that the link is congested, the packet belongs to If the destination IP address #j is not entered in the table, the packet is discarded. If the destination IP address #j is already entered in the table, the counter value C drop ( j) is read from the table and the packet is discarded with probability P dec (C drop (j)) (where p dec (x) is 0 ≦ P pdec (x) ≦ 1 and monotonically decreasing function with respect to x)) It is characterized by doing.
In the method of the fourth embodiment, the number of accepted SYN packets is managed for each user. However, in the method of the fifth embodiment, the number of discarded SYN packets is managed, and the number of discarded users is large. By reducing the discard rate, fairness among users is ensured.

In the TCP traffic control method according to the sixth embodiment of the present invention, if the in-buffer packet length Q in the output waiting buffer exceeds a predetermined threshold, it is determined that the link is congested, It is characterized in that selective discarding of packets or priority control is performed by any one of methods 1 to 5.
In the TCP traffic control method according to the seventh embodiment of the present invention, the number of arrival bytes A at the link is measured every predetermined period t0, and when the packet arrives at the node, the current number of arrival bytes If A exceeds a predetermined threshold, it is determined that the link is congested, and packet selective discarding or priority control is performed by any one of the first to fifth embodiments. It is characterized by.
In the TCP traffic control method according to the eighth embodiment of the present invention, the arrival number Asyn of the SYN flagged packets to the link is measured every predetermined period t0, and when the packet arrives at the node, If the arrival number Asyn of the SYN flagged packet exceeds the predetermined threshold value, it is determined that the link is congested, and the packet is selectively discarded by any one of the first to fifth embodiments. Alternatively, priority control is performed.

In the TCP traffic control method according to the ninth embodiment of the present invention, when a packet arrives at the node, the TCP header of the packet is read, and if the SYN flag indicating the start of the TCP connection is set, the counter L is set to 1. If the FIN flag or RST flag indicating the end of the count up or TCP connection is set, L is counted down to know the number L of simultaneous connection connections, and when a packet arrives at the node, the current number of simultaneous connection connections If L exceeds a predetermined threshold, it is determined that the link is congested, and selective packet discard or priority control is performed by any of the methods of the first to fifth embodiments. It is characterized by.
In the TCP traffic control method according to the tenth embodiment of the present invention, as in the method of the eighth embodiment, the arrival number Asyn of the packets with the SYN flag on the link is measured every predetermined period t0. In addition, the number L of simultaneous connections is measured in the same manner as in the ninth embodiment, and the average value E [L] is calculated every period t0, and the average file transfer is performed from Asyn and E [L]. The time E [T] is estimated by E [T] = E [L] / (Asyn / t0), and when a packet arrives at the node, the current E [T] exceeds a predetermined threshold. Then, it is judged that the link is congested, and the packet is selectively discarded by any one of the first to fifth embodiments.
In the TCP traffic control apparatus according to the eleventh embodiment of the present invention, means for performing selective packet discard or priority control by any of the methods of the first to fifth embodiments, and The present invention is characterized by comprising means for grasping a link congestion state by any one of the methods of the tenth embodiment.

  According to the present invention, it is possible to effectively use network resources even during network congestion, avoid the need for special signaling protocol and state management for each connection, and avoid continuation and promotion of congestion and invalid suspension of resources. It is possible to ensure the TCP quality during communication.

Embodiments of the present invention will be described below in detail with reference to the drawings.
(First embodiment)
FIG. 1 is a basic configuration diagram of an IP network to which the present invention is applied.
The present invention is applied by providing a TCP traffic control device at the ingress node 10 in FIG. 20 is an outgoing node, and 30 is a relay link.
FIG. 2 is a block diagram of a TCP traffic control apparatus used in the packet selective discard method according to the first embodiment of the present invention.
The structure of the link congestion determination method according to the sixth embodiment, which will be described later, and the TCP traffic control apparatus according to the eleventh embodiment are also described with reference to FIG.
As shown in FIG. 2, the packet arriving from the user is transferred to the header analysis unit 41, the traffic monitoring unit 44, and the packet discard determination unit 42. The packet transfer unit 43 outputs the processed packet to the link.
The header analysis unit 41 checks whether the SYN flag indicating the start of the TCP connection is set in the TCP header of the packet, and notifies the packet discard determination unit 42 of the result.

On the other hand, the traffic monitoring unit 44 that has received a packet from the user receives the current queue length Q.
[Packets] is updated by Q ← Q + 1, and the packet discard determining unit 42 is notified of the value of Q. The packet discard determination unit 42 compares the value of Q with a predetermined threshold value Th, and discards the packet if Q> Th and the SYN flag is set in the packet. The traffic monitoring unit 44 updates the queue length by Q ← Q-1. Otherwise, the packet is accepted and transferred to the packet transfer unit 43. The packet transfer unit 43 that has received the packet accumulates the packet in the buffer. The packet is taken out from the head of the buffer at the output link speed. If the packet is taken out, the traffic monitoring unit 44 is notified accordingly. The traffic monitoring unit 44 updates the queue length Q by Q ← Q-1.
Note that instead of directly using Q as the queue length in the traffic monitoring unit 44, an average queue length may be used. The average queue length is set to Qave, and is updated and held by Qave ← (1−a) × Qave + a × Q (a is a smoothing parameter) every fixed period t0.

(Second embodiment)
FIG. 3 is a block diagram of a control device of the TCP traffic control method according to the second embodiment of the present invention.
In the second embodiment of the present invention, instead of providing the packet discard determining unit 42 as in the first embodiment, the packet forwarding unit 43 uses two classes (high priority class, low Priority class is performed by extracting packets from the class buffers 45 and 46 according to a predetermined weight.
When a packet arrives at a node, if a SYN flag indicating the start of a TCP connection is set in the TCP header of the packet, the packet is accommodated in the low priority class buffer 46, otherwise, the high priority class buffer 45 And priority control.

(Third embodiment)
A third embodiment of the present invention will be described with reference to FIGS.
As in the first and second embodiments, the header analysis unit 41 checks whether the SYN and ACK flags are set instead of checking whether the SYN flag is set. In other words, in FIG. 2, when a packet from the user arrives, the header analysis unit 41 determines whether or not the SYN flag indicating the start of the TCP connection is set and the ACK flag is set. The header analysis unit 41 notifies the packet discard determination unit 42 so as to discard only the packets flagged in both of them.
Also in FIG. 3, when the packet from the user arrives, the header analysis unit 41 determines whether or not both the SYN flag and the ACK flag are set, and the packet with both flags set is determined. The low priority class buffer 46 is accommodated, and other packets are accommodated in the high priority class buffer 45. The packet transfer unit 43 extracts the packet with priority from the high priority class buffer 45 and outputs it to the link.

(Fourth embodiment)
FIG. 4 is a table configuration diagram used in the TCP traffic control method according to the fourth embodiment of the present invention, and FIG. 5 is an internal configuration of the header analysis unit used in the TCP traffic control method according to the fourth embodiment. FIG.
In this embodiment, a table for managing the destination IP address of the user who has transmitted a packet flagged in SYN and ACK is prepared, and when a packet flagged in SYN and ACK arrives, If the packet is accepted without being discarded by the method described later, the destination IP address #i of the packet is read, and if the IP address #i has already been entered in the table, the SYN packet passage counter C pass (I) is incremented by 1, and if it is a new packet not entered in the table, it is added to the table and the counter C pass (i) is set to 1. When a predetermined timeout time Tout has elapsed since the counter was last updated, C pass (i) ← INT [C pass (i) × α] (α is a parameter satisfying 0 <α <1, INT [ x] is the largest integer not exceeding x), and when C pass (i) = 0, the address #i is deleted from the table.

As shown in FIG. 4, the table 412 stores, for example, the arrival number of the packet, the SYN in the TCP header, the presence / absence of the ACK flag, and the destination IP address.
As shown in FIG. 5, both the table 412 and the SYN passage counter 411 are arranged in the header analysis unit 41, and the destination IP address of the table 412 is referred to under the control of the control circuit 413 in the header analysis unit. , The SYN passage counter 411 is incremented or decremented.
When the packet from the user arrives at the node and the destination IP address of the arrival packet is j, it is confirmed that the SYN and ACK in the TCP header of the packet are flagged and the link is congested. If detected, if the destination IP address #j to which the packet belongs is not entered in the table 412, the packet is accepted. If the destination IP address #j has already been entered in the table 412, the IP address is accepted. The counter value C pass (j) corresponding to #j is read from the table 412 and the probability P inc (C pass (j)) (where P inc (x) is 0 ≦ P inc (x) ≦ 1) , X with a monotonically increasing function). That is, as shown in FIG. 2, the header analysis unit 41 notifies the packet discard determination unit 42 so that the packet is discarded.

(Fifth embodiment)
FIG. 6 is a configuration diagram of the header analysis unit used in the TCP traffic control method according to the fifth embodiment of the present invention.
A table 412 for managing the destination IP address of a user who has transmitted a packet flagged in SYN and ACK is prepared, and when a packet flagged in SYN and ACK arrives, If the packet is discarded, the destination IP address #i of the packet is read. If the IP address #i has already been entered in the table 412, the SYN packet discard counter C drop (i) 414 is incremented by one. If the packet is a new packet not entered in the table 412, it is added to the table 412 and the counter 414 is set to 1. When the predetermined timeout time Tout has elapsed since the counter 414 was last updated, the counter C drop (i) ← INT [C drop (i) × α] (where α is 0 <α <1 Parameter [INT [x] is the largest integer not exceeding x), and when C drop (i) = 0, the address #i is deleted from the table.

That is, instead of managing the number of SYN packets accepted without being discarded as in the fourth embodiment, the traffic monitoring unit 44 manages the number of dropped SYN packets C drop (j) and decides to discard the packet. In the unit 42, instead of determining the discard probability with P inc (C pass (j)) as in the fourth embodiment, the probability P dec (C drop (j)) (P dec (x) is 0 ≦ P dec (x) ≦ 1, and the packet is discarded with a monotonically decreasing function with respect to x).
In FIG. 2, when a packet whose destination IP address is j arrives at a node, the header analysis unit 41 is flagged in SYN and ACK in the TCP header of the packet, and the link is congested. If the destination IP address #j to which the packet belongs is not entered in the table 412, the packet discard determining unit 42 is notified to discard the packet. On the other hand, if the destination IP address #j has already been entered in the table 412, the value C drop (j) of the counter 414 corresponding to the IP address #j is read from the table 412 and the probability P dec (C drop ( j)) (where P dec (x) is 0 ≦ P dec (x) ≦ 1 and is a monotonically decreasing function with respect to x), the packet discard determining unit 42 is notified to discard the packet. Note that reference numeral 442 in FIG. 6 denotes a counter that measures the queue length Q of the buffer.

(Sixth embodiment)
A sixth embodiment of the present invention will be described.
When the packet length Q in the buffer in the output waiting buffer exceeds a predetermined threshold, it is determined that the link is congested, and the packet is selected by any of the methods of the first to fifth embodiments. Implement disposal or priority control.
In this embodiment, a timer is further arranged. That is, instead of monitoring the queue length in the traffic monitoring unit 44, the number of arrival bytes A to the link is measured every predetermined period t0, and it is congested when it exceeds a predetermined value. It is determined, and it is determined whether or not packet discard is performed using the determination.
The header analysis unit 41 determines that the link is congested when notified from the traffic monitoring unit 44 that the in-buffer packet length Q in the output waiting buffer exceeds a predetermined threshold value, and determines that the link is congested. The packet selective discard or priority control is performed by any of the methods from the embodiment to the fifth embodiment.

  That is, the sixth embodiment is a method of measuring whether or not the packet length Q exceeds a threshold value as a criterion for determining whether or not there is congestion. For example, in the first embodiment, as shown in FIG. 2, when the header analysis unit 41 is notified from the traffic monitoring unit 44 that the packet length Q exceeds the threshold value, the TCP of the transmitted packet is transmitted. If it is detected that the SYN flag is set in the header, the packet discard determining unit 2 is notified of the fact and the packet is discarded. In the second embodiment, as shown in FIG. 3, when the header analysis unit 41 detects that the packet length Q exceeds the threshold, the SYN flag is set in the TCP header of the packet sent. When it is detected that the packet is standing, the packet is accommodated in the buffer 46 of the low priority class, and when it is not detected that the SYN flag is standing, the packet is accommodated in the buffer 45 of the high priority class.

(Seventh embodiment)
Instead of monitoring the queue length by the traffic monitoring unit 44 in the first embodiment, the number of arrival bytes A to the link is measured every predetermined period t0, and congestion is detected when a predetermined threshold is exceeded. judge. Using this, it is determined whether or not to discard the packet or perform priority control by any of the methods of the first to fifth embodiments. For example, as shown in FIG. 7, the traffic monitoring unit 44 sets the timer 444 to measure the number of bytes A of packets arriving on the link every predetermined period t0 by the counter 441.

(Eighth embodiment)
In the eighth embodiment, instead of the queue length, the arrival number Asyn of the packets with the SYN flag is measured as a criterion for determining whether or not there is congestion, and it is used to determine whether or not to discard the packet. . Therefore, the methods of the first to fifth embodiments are combined.
For example, as shown in FIG. 7, instead of monitoring the queue length by the traffic monitoring unit 44, the control circuit 443 determines whether or not the SYN flag is set in the TCP header of the packet that has arrived. Then, the arrival number Asyn of the packets with the SYN flag is measured every predetermined period t0, and if it exceeds the predetermined value, it is determined as congestion, and it is determined whether or not to discard the packet.

  When the first embodiment is combined, when the packet arrival number Asyn exceeds a predetermined threshold, the packet discard determination unit 42 is notified of all packets with the SYN flag and discarded. When the second embodiment is combined, when the packet arrival number Asyn exceeds the threshold, the packet with the SYN flag is accommodated in the low priority class buffer 46, and the other packets are accommodated in the high priority class buffer 45. . When the third embodiment is combined, when the packet arrival number Asyn exceeds the threshold, only the packet for which the ACK flag is set in addition to the SYN flag is discarded or accommodated in the low priority class buffer 46. When the fourth embodiment is combined, when the packet arrival number Asyn exceeds the threshold, if the packet with the SYN flag and the ACK flag has already been entered in the table, the probability P inc (C pass (j)) If it is discarded and newly entered, it is accepted. When the fifth embodiment is combined, when the packet arrival number Asyn exceeds the threshold, the number of dropped SYN packets C drop (j) is managed, and the packet with the probability P dec (C drop (j)) Discard.

(Ninth embodiment)
FIG. 8 is a block diagram of a traffic monitoring unit according to the ninth embodiment of the present invention.
The traffic monitoring unit 44 in the first embodiment monitors the queue length. In this embodiment, if the SYN flag indicating the start of the TCP connection is set instead, the simultaneous connection connection counter L is set to 1. If the FIN flag or the RST flag indicating the end of the TCP connection is set up, the counter L is decremented by 1 to grasp the number L of simultaneously connected connections, and the value L exceeds the predetermined value. If it is determined that the packet is discarded, it is determined whether or not to discard the packet.

  When the first embodiment is combined, when the number L of simultaneously connected connections exceeds a predetermined threshold, the packet discard determining unit 42 is notified of all packets with the SYN flag and discarded. When the second embodiment is combined, when the number L of simultaneous connections exceeds the threshold, packets with the SYN flag are accommodated in the low priority class buffer 46, and other packets are accommodated in the high priority class buffer 45. To do. When the third embodiment is combined, when the number L of simultaneously connected connections exceeds the threshold, only the packet for which the ACK flag is set in addition to the SYN flag is discarded or accommodated in the low priority class buffer 46. When the fourth embodiment is combined, when the number of simultaneously connected connections L exceeds the threshold, if the packet with the SYN flag and the ACK flag has already been entered in the table, the probability P inc (C pass (j)) If it is discarded and newly entered, it is accepted. When the fifth embodiment is combined, the number of dropped SYN packets C drop (j) is managed and the probability P dec (C drop (j)) when the number L of simultaneous connections exceeds the threshold. Discard the packet.

(Tenth embodiment)
The tenth embodiment of the present invention is a combination of the eighth and ninth embodiments.
That is, instead of monitoring the queue length in the traffic monitoring unit 44 in the first embodiment, the arrival number Asyn of the packets with the SYN flag to the link is determined every predetermined period t0, as in the eighth embodiment. In the same manner as in the ninth embodiment, the number L of simultaneous connections is measured, the average value E [L] is calculated every period t0, and the average is calculated from Asyn and E [L]. The file transfer time E [T] is estimated by E [T] = E [L] / (Asyn / t0), and if it exceeds a predetermined value, it is determined as congestion. It is used to decide whether to discard the packet.

(Eleventh embodiment)
FIG. 2 is a block diagram of a TCP traffic control apparatus according to the eleventh embodiment of the present invention.
Means for performing selective discard or priority control of packets by any of the methods of the first to fifth embodiments (header analysis unit 41 and packet discard determination unit 42 in FIG. 2), means for grasping link congestion status (FIG. 2 traffic monitoring unit 44) and means for transferring the processed packet to the link (packet transfer unit 43 in FIG. 2).

It is a figure which shows the basic composition of the IP network to which this invention is applied. It is a block diagram of the TCP traffic control apparatus which concerns on one Example of this invention. It is a block diagram of the TCP traffic control apparatus based on 2nd Example of this invention. It is a block diagram of the table which manages the destination IP address based on the 4th Example of this invention. It is a block diagram of the header analysis part which concerns on the 4th Example of this invention. It is a block diagram of the header analysis part which concerns on the 5th Example of this invention. It is a block diagram of the traffic monitoring part which concerns on the 7th and 8th Example of this invention. It is a block diagram of the traffic monitoring part which concerns on the 9th Example of this invention.

Explanation of symbols

10 ... Incoming node, 20 ... Outgoing node, 30 ... Relay link, 40 ... Node,
41 ... Header analysis unit, 42 ... Packet discard determination unit, 43 ... Packet transfer unit,
44 ... Traffic monitoring unit, 45 ... High priority class buffer, 46 ... Low priority class buffer,
412 ... Destination IP address management table, 411 ... SYN packet passage counter,
413 ... Control circuit, 414 ... SYN packet discard counter,
441 ... SYN packet passage counter, 444 ... timer, 443 ... control circuit,
444... Counter, 445.

Claims (11)

In a TCP traffic control method when a TCP connection is multiplexed on a link between nodes in a communication network,
When a packet arrives at the node, if the SYN flag indicating the start of a TCP connection is set in the TCP header of the packet and it is detected that the link is congested, the packet is discarded. A TCP traffic control method characterized by: In a TCP traffic control method when a TCP connection is multiplexed on a link between nodes in a communication network,
A control method for performing priority control by dividing an output waiting buffer of the link into two classes and extracting packets from each class buffer according to a predetermined weight,
When a packet arrives at the node, if the SYN flag indicating the start of the TCP connection is set in the TCP header of the packet and it is detected that the link is congested, the packet is reduced. A TCP traffic control method characterized in that priority control is performed by accommodating in a priority class buffer and otherwise accommodating in a high priority buffer. In the TCP traffic control method according to claim 1 or 2,
When a packet arrives at the node, only the packet flagged in both the SYN indicating the start of the TCP connection and the ACK indicating the completion of the packet reception in the TCP header of the packet is selectively discarded or A TCP traffic control method, wherein priority control is performed. In a TCP traffic control method when a TCP connection is multiplexed on a link between nodes in a communication network,
When a packet with a flag set in SYN and ACK is transmitted from the user terminal, a table for managing the destination IP address is arranged,
When a packet whose destination IP address is j arrives at the node, if SYN and ACK are flagged in the TCP header of the packet and it is detected that the link is congested, If the destination IP address j to which it belongs is not entered in the table, it accepts the packet, adds it to the table,
If the destination IP address j is already entered in the table, the value of the SYN packet passage counter corresponding to the IP address j is read from the table, and the probability P inc (C pass (j)) (where P inc (x) is 0 ≦ P inc (x) ≦ 1, and the packet is discarded with a monotonically increasing function with respect to x). In a TCP traffic control method when a TCP connection is multiplexed on a link between nodes in a communication network,
When a packet with a flag set in SYN and ACK is transmitted from the user terminal, a table for managing the destination IP address is arranged,
When a packet whose destination IP address is j arrives at the node, if SYN and ACK are flagged in the TCP header of the packet and it is detected that the link is congested, the packet If the destination IP address j to which it belongs is not entered in the table, the packet is discarded,
If the destination IP address j is already entered in the table, the value (C drop (j)) of the SYN packet discard counter corresponding to the IP address j is read from the table, and the probability P dec (C drop (j )) (Where p dec (x) is 0 ≦ P pdec (x) ≦ 1 and is a monotonically decreasing function with respect to x), the TCP traffic control method characterized by discarding the packet. In the TCP traffic control method according to any one of claims 1 to 5,
If the in-buffer packet length Q in the output waiting buffer of the node exceeds a predetermined threshold value, it is determined that the link is congested, and selective discard or arrival control of arrival packets is performed. TCP traffic control method. In the TCP traffic control method according to any one of claims 1 to 5,
In the node, the number of arrival bytes A to the link is measured every predetermined period t0, and when a packet arrives at the node, the current number of arrival bytes A exceeds a predetermined threshold. A TCP traffic control method characterized by judging that the link is congested and performing selective discard or priority control of the packet. In the TCP traffic control method according to any one of claims 1 to 5,
The node measures the arrival number Asyn of the packets with the SYN flag on the link every predetermined period t0, and when the packet arrives at the node, the arrival number Asyn of the current packet with the SYN flag is predetermined. A TCP traffic control method characterized in that if the threshold is exceeded, it is determined that the link is congested, and selective discard or priority control of the packet is performed. In the TCP traffic control method according to any one of claims 1 to 5,
In the node, when the packet arrives at the node, the TCP header of the packet is read, and if the SYN flag indicating the start of the TCP connection is set, the simultaneous connection connection counter L is incremented by 1, and the FIN indicating the end of the TCP connection is determined. If the flag or the RST flag is set, L is counted down to know the number L of simultaneous connections,
When a packet arrives at the node, if the number L of current simultaneous connection connections exceeds a predetermined threshold, it is determined that the link is congested, and selective discard or priority control of the packet is performed. And a TCP traffic control method. In the TCP traffic control method according to any one of claims 1 to 5,
The node measures the arrival number Asyn of packets with a SYN flag to the link every predetermined period t0, and measures the number L of simultaneous connections.
The average value E [L] of the number L of simultaneously connected connections is calculated every period t0,
The average file transfer time E [T] is estimated from Asyn and E [L] by E [T] = E [L] / (Asyn / t0),
When a packet arrives at the node, if the current E [T] exceeds a predetermined threshold, it is determined that the link is congested, and selective discard of the packet is performed. TCP traffic control method. In a TCP traffic control device when a TCP connection is multiplexed on a link between nodes in a communication network,
A TCP traffic control apparatus comprising means for performing selective discard or priority control of packets, and means for grasping the congestion status of the link and notifying the congestion status to the means for performing selective discard or priority control .

Images