JP2018033119A - Packet communication system and congestion control method therefor, and congestion control program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a packet communication system and congestion control method, capable of achieving a proper response to rapid variations in communication quality such as a bandwidth.SOLUTION: The packet communication system, having a radio base station 200 storing a terminal and a communication device 400 forming a connection of a transport layer of an OSI7 hierarchy model between the terminal and itself, acquires radio access quality between the terminal and the radio base station 200 and adjusts a congestion window size in the transport layer on the basis of the radio access quality using a congestion window control unit 403.SELECTED DRAWING: Figure 5

Description

  The present invention relates to a packet communication system, and more particularly to a congestion control function of a transport layer.

  In current Internet communication, TCP / IP (Transmission Control Protocol / Internet Protocol) protocol is used as standard. TCP, which is a typical communication protocol in the transport layer, is a connection-type reliability protocol, and has retransmission control and congestion control functions. The basic operation of TCP is that the receiving side returns an acknowledgment (ACK) indicating that the packet has been received in response to the packet sent by the transmitting side, and the transmitting side that has received ACK further transmits the next packet. At this time, network congestion control is performed by adjusting a congestion window (cwnd) indicating the amount of data that can be transmitted without waiting for an ACK. In order to avoid congestion at the start of communication, cwnd performs a slow start that increases cwnd by one packet each time an ACK is received. When packet loss occurs due to congestion or when cwnd exceeds a slow start threshold (ssthresh), a transition is made to a congestion avoidance mode in which control is performed to increase or decrease cwnd so as to avoid congestion.

  Various methods have been proposed for congestion control algorithms as Internet communication changes. CUBIC (see Non-Patent Document 1), which is standard in Linux kernel 2.6 and later versions, has an RTT (Round Trip Time) fairness because the amount of increase in cwnd in the congestion avoidance mode depends on the time from packet loss. Communication throughput can be improved even in a high and high delay environment. In recent years, wireless communication has become common, but wireless communication has a large variation in communication quality compared to wired communication, and there is a problem of reduced throughput due to increased delay and packet loss due to wireless errors.

  As a technique for improving the throughput in wireless communication, TCP parameters and options such as Westwood, which controls the prevention of unnecessary throughput reduction by estimating the available bandwidth at the time of packet loss, expansion of maximum cwnd, use of SACK, etc. There are W-TCP (Wireless-Profiled TCP) tuned for wireless communication.

  In addition, a cross-layer control method for performing TCP control by notifying not only the transport layer but also the Internet layer and the data link layer of the congestion status and wireless handover has been studied.

  Quick-Start (see Non-Patent Document 2) notifies the requested throughput on the transmission side using the option area of the TCP and IP packets, and responds to the requested throughput if all nodes between transmission and reception approve. This is a method of changing to cwnd.

  Further, Non-Patent Document 3 introduces a technique for notifying a server of a throughput that can be explicitly used by a change in a radio link or the like using an option area of a TCP packet.

  A method for detecting communication quality and switching a plurality of congestion control functions based on the communication quality has also been proposed.

Injong Rhee, and Lisong Xu, "CUBIC: A New TCP-Friendly High-Speed TCP Variant", 2008 S. Floyd, 3 others, "Quick-Start for TCP and IP", RFC4782 A. Jain et al., "Mobile Throughput Guidance Inband Signaling Protocol draft-flinck-mobile-throughput-guidance-03", IETF Internet-Draft, 2015

  The wireless network of the 2020s may be a multi-RAT environment in which a plurality of RATs (Radio Access Technology) having different characteristics such as a use frequency band and a cell radius are mixed to efficiently accommodate increasing traffic. Assumed. As shown in FIG. 1, in addition to the existing LTE (Long Term Evolution) and Wi-Fi, the RAT to be used is, for example, a maximum band of several G to several tens of Gbps by using a high-frequency millimeter wave. A cheaper one with a small radius or an unlicensed band such as a 5 GHz band but low reliability can be considered. In such a Multi-RAT environment, small cells with a cell radius of several meters to several hundred meters are scattered, so that switching of cells due to movement of communication terminals frequently occurs, and bandwidth, delay, and PER (Packet Error Rate). Etc. The communication quality fluctuates rapidly.

  CUBIC, the TCP congestion control algorithm described above, is a method that improves the throughput in a broadband and high-delay environment while improving fairness with RTT and other TCP flows, and was designed to adapt to rapid changes in the communication environment. It is not a thing. For this reason, when connected to a cell using an ultra-wideband millimeter wave with a narrow cover range during communication by CUBIC in LTE, the cwnd is not increased rapidly in order to use the free band. Therefore, there is a problem that if the band is moving, the band is taken out before the band is fully used and the band cannot be fully utilized.

  In addition, CUBIC determines that packet loss is congestion and lowers throughput, so when using an unlicensed band, there is a problem that PER increases due to radio wave interference and unnecessarily lowers throughput.

  Further, the requested throughput can be obtained immediately by using Quick-Start, but in order to use Quick-Start, all nodes between the transmitting and receiving terminals need to support Quick-Start.

  In addition, switching of the congestion control function based on the communication quality described above requires passing parameters between the congestion control functions, synchronizing control switching timing between the transmitting and receiving terminals, and the like.

  The present invention has been made in view of the above circumstances, and provides a packet communication system, a congestion control method thereof, and a congestion control program capable of quickly responding to rapid fluctuations in communication quality such as bandwidth. .

  In order to achieve the above object, the present invention provides a packet communication system including a radio base station that accommodates radio terminals and a communication device that forms a transport layer connection of the OSI7 hierarchical model with the radio terminals. Radio access quality detection means for detecting radio access quality between a radio terminal and a radio base station, and a congestion window for adjusting a congestion window size in the transport layer based on the radio access quality detected by the radio access quality detection means And a control means.

  According to the present invention, by performing congestion window control based on the quality change of the radio access environment, for example, it is detected that the band is widened by applying carrier aggregation or dual connectivity during communication by LTE, and the broadband is immediately It becomes possible to change to the control to be used. Further, when moving to a narrow band cell again, fairness with other flows can be maintained by controlling so that cwnd is not increased rapidly.

  Also, by considering the radio transmission quality in the transport layer, it is possible to determine the increase in RTT and PER due to radio errors and congestion, and it is possible to avoid unnecessarily reducing the throughput.

Image diagram explaining the Multi-RAT environment Configuration diagram of packet communication system via proxy Configuration diagram of packet communication system without proxy Functional configuration diagram of Embodiment 1 Sequence diagram of Embodiment 1 Functional configuration diagram of Embodiment 2 The figure which shows an example of the maximum bandwidth conversion table Sequence diagram of embodiment 2 Functional configuration diagram of Embodiment 3 Operation flow of wireless transmission quality fluctuation judgment unit Other operation flow of wireless transmission quality fluctuation judgment unit Sequence diagram of Embodiment 3 Functional configuration diagram of Embodiment 4 Sequence diagram of embodiment 4 Functional configuration diagram of Embodiment 5 Operation flow of congestion window control 1 Operation flow of congestion window control 2 Graph showing cwnd increase in each mode Operation flow of congestion window control 3 Operation flow of congestion window control 4 Existing Transport Layer Congestion Window Controller Operation flow of congestion window control 5 The graph which shows the cwnd increase / decrease amount in the congestion window control 6 The graph which shows the change of the cwnd increase / decrease amount by the control which combined the congestion window control 1 and 6

  FIG. 2 shows a packet communication system for implementing the present invention. The communication system includes a terminal 100, a base station 200, a core node 300 that accommodates the base station, a proxy 400 that is a communication device that terminates an OSI (Open Systems Interconnection) 7-layer model transport layer connection, and a communication destination of the terminal 100. The server 100 includes a certain server 500, and the terminal 100 communicates with the server 500 via the base station 200, the core node 300, and the proxy 400. The deployment location of the proxy 400 is not limited to the connection point of the Internet, but may be considered in the core network. As another embodiment, as shown in FIG. 3, the terminal 100 can directly communicate with the server 500 without going through the proxy 400. In this case, the server 500 is a communication device that terminates the transport layer connection.

  FIG. 21 shows an existing transport layer congestion window control unit. Since the existing congestion window control unit does not acquire information from outside the transport layer, the information that can be used for congestion control is limited to the amount of packet transmission, the amount of ACK received, the round trip time (RTT) of the packet, and the like.

  In the present invention for congestion window control in the existing transport layer, (1) the terminal 100 or the base station 200 calculates the radio access quality, and the transmission side communication device (terminal 100 or proxy 400 or server 500) (2) Control for adjusting the TCP congestion window size cwnd based on the radio access quality in the transport layer control unit of the transmission side communication device (terminal 100 or proxy 400 or server 500) It is characterized by performing.

  The radio access quality is, for example, the maximum bandwidth that can be used by the terminal 100 (hereinafter simply referred to as the maximum bandwidth), the radio transmission quality between the terminal 100 and the base station 200 (eg, Received Signal Strength Indicator (RSSI), SINR). (Signal to Interference Noise Ratio), CQI (Channel Quality Indicator) in LTE, etc.), etc., are indicators indicating the quality of the radio access environment.

  There are the following five embodiments for notifying the quality of the radio access environment (1) of the present invention.

[Embodiment 1] Mode of Notifying Maximum Bandwidth from Base Station 200 to Proxy 400 [Embodiment 2] Mode of Notifying Maximum Bandwidth from Terminal 100 to Proxy 400 [Third Embodiment] Radio Transmission Quality from Base Station 200 to Proxy 400 [Embodiment 4] Mode of Notifying Wireless Transmission Quality from Terminal 100 to Proxy 400 [Embodiment 5] Mode of Performing TCP Congestion Window Control in Terminal 100 Hereinafter, each embodiment will be described in detail.

[Embodiment 1]
Embodiment 1 is a form in which the base station 200 notifies the proxy 400 of the maximum bandwidth that the terminal 100 can use. The functions of the base station 200 and the proxy 400 of the first embodiment are shown in FIG. The base station 200 calculates the maximum bandwidth that can be used by the terminal 100 from the frequency band that can be allocated to the terminal 100, the modulation method, the presence / absence of MIMO (Multiple-Input Multiple-Output) use, the number of communication flows that share a wireless link, and the like. The maximum bandwidth calculation unit 201 is provided. The proxy 400 includes, in the transport layer, an RTT measurement unit 401 that measures RTT for each packet transmission / reception, an access quality monitoring unit 402 that detects a change in radio access quality, an RTT and an access quality monitoring unit acquired from the RTT measurement unit 401 It has a congestion window control unit 403 that reflects it in TCP congestion window control based on the maximum bandwidth acquired from 402. FIG. 5 shows a sequence diagram of the first embodiment.

  As shown in FIG. 5, when the maximum bandwidth calculation unit 201 detects a change in the maximum bandwidth, it notifies the access quality monitoring unit 402 of the maximum bandwidth. The congestion window control unit 403 acquires the RTT from the RTT measurement unit 401, acquires the maximum bandwidth from the access quality monitoring unit 402, and controls the congestion window based on the RTT and the maximum bandwidth.

  When the terminal 100 communicates with the server 500 without going through the proxy 400 (see FIG. 3), the server 500 has the function of the proxy 400.

[Embodiment 2]
In the second embodiment, the maximum bandwidth is notified from the terminal 100 to the proxy 400. The functions of the terminal 100 and the proxy 400 of the second embodiment are shown in FIG. The terminal 100 has a wireless communication management unit 101 that manages the type, frequency band, and MIMO usage of the wireless access being used, a maximum bandwidth conversion table 102 that converts the maximum bandwidth from the communication environment being used, and the maximum bandwidth has changed. In this case, a wireless quality notification unit 103 that notifies the access quality monitoring unit 402 of the proxy 400 of the maximum bandwidth is provided. The wireless quality notification unit 103 notifies the proxy 400 side by adding maximum bandwidth information to the option area or payload part of the header part of the TCP ACK packet. The proxy 400 has the same function as that of the first embodiment, and performs the same TCP congestion window control as that of the first embodiment in the congestion window control unit 403 when the maximum bandwidth changes.

  The maximum bandwidth conversion table 102 is shown in FIG. In the table of FIG. 7, the terminal 100 cannot know the exact maximum bandwidth, but it is an advantage of the second embodiment that the present invention can be implemented without modifying the base station.

  The procedure after the notification of the maximum bandwidth to the access quality monitoring unit 402 is the same as in the first embodiment. FIG. 8 shows a sequence diagram of the first embodiment.

  When the terminal 100 communicates with the server 500 without going through the proxy 400 (see FIG. 3), the server 500 has the function of the proxy 400.

[Embodiment 3]
The third embodiment is a form in which the radio transmission quality is notified from the base station 200 to the proxy 400. The functions of the base station 200 and the proxy 400 of the third embodiment are shown in FIG. The terminal 100 includes a radio communication management unit 101 that detects radio wave reception quality from the base station 200 as radio transmission quality, and a radio quality notification unit 103 that periodically notifies the detected radio transmission quality to the base station 200. As the radio transmission quality, in addition to RSSI, SINR, etc., LTE uses CQI. The base station 200 includes a wireless transmission quality acquisition unit 202 that acquires wireless transmission quality from the wireless quality notification unit 103, and a wireless transmission quality variation determination unit 203 that determines a variation in wireless transmission quality of the terminal 100 from the wireless transmission quality acquired a plurality of times. It comprises. When RSSI and SINR are used, class values obtained by classifying into several classes (for example, classes 1 to 15) representing good quality from RSSI and SINR values are used. Since the radio transmission quality changes every moment according to the usage status of the terminal 100, the process of determining the radio transmission quality variation such as the average value of the radio transmission quality for a plurality of times or the process of assuming that the radio transmission quality has changed when a certain number of times exceeds the reference value is determined. This is performed in the unit 203.

An example of the operation of the wireless transmission quality variation determination unit 203 is shown in FIGS. The example of FIG. 10 calculates an average value of wireless transmission quality (Q ₁ (in FIG. 10, an overline is added to represent “average”. The same applies hereinafter)) at regular intervals. Steps S101, S102), the quality fluctuates if the difference between the previously calculated average value (Q ₁ ) and the newly calculated average value (Q ₂ ) of the radio transmission quality exceeds a reference value that the radio transmission quality is considered to have fluctuated. It is determined that it has been performed (steps S103 to S107). In the example of FIG. 11, two reference values 1 and 2 are determined, and if the acquired wireless transmission quality is greater than the reference value 1, it is determined that the quality is better than the current wireless transmission quality of the terminal 100 (steps S111 to S113). If it is smaller than the reference value 2, it is determined that the quality is lower than the current transmission quality of the terminal 100 (steps S114 to S115). If it is determined that the quality is good or bad for a certain number of times within a certain period, it is determined that the wireless transmission quality has changed (steps S116 to S117).

  The proxy 400 has the same functions as those in the first and second embodiments. When the wireless transmission quality variation determining unit 203 determines that the wireless transmission quality has changed, the access quality monitoring unit 402 displays the changed wireless transmission quality. Notice. The access quality management unit 402 notifies the congestion window control unit 403 of the wireless transmission quality, and changes the TCP congestion window control. A sequence diagram of the third embodiment is shown in FIG.

  When the terminal 100 communicates with the server 500 without going through the proxy 400 (see FIG. 3), the server 500 has the function of the proxy 400.

[Embodiment 4]
The fourth embodiment is a form in which the wireless transmission quality is notified from the terminal 100 to the proxy 400. The functions of the terminal 100 and the proxy 400 of the fourth embodiment are shown in FIG. The terminal 100 includes a wireless communication management unit 101 that detects wireless transmission quality from communication with the base station 200, a wireless transmission quality variation determination unit 104 that determines a variation in quality from a plurality of times of wireless transmission quality, and a wireless transmission quality that has changed. A wireless quality notification unit 103 that notifies the proxy 400 at this time. The proxy 400 includes the same function unit as that of the third embodiment. The wireless transmission quality variation determination unit 104 performs the same processing (FIGS. 10 and 11) as the wireless transmission quality variation determination unit 203 according to the third embodiment. That is, the fourth embodiment is characterized in that the radio transmission quality variation determination performed by the base station 200 in the third embodiment is performed on the terminal 100 side. The notification of the wireless transmission quality from the wireless quality notification unit 103 to the proxy 400 is performed by adding the wireless transmission quality information to the option area or the payload part of the header part of the TCP ACK packet as in the second embodiment. To notify. FIG. 14 shows a sequence diagram of the fourth embodiment.

  When the terminal 100 communicates with the server 500 without going through the proxy 400 (see FIG. 3), the server 500 has the function of the proxy 400.

[Embodiment 5]
In the fifth embodiment, TCP congestion window control is performed in the terminal 100. The functions of the fifth embodiment are shown in FIG. In the first to fourth embodiments, the proxy 400 is the transmission side, but in the fifth embodiment, the terminal 100 side is the transmission side, and the terminal 100 includes the RTT measurement unit 105 and the congestion window control unit 106. The function of the access quality monitoring unit 401 of the proxy 400 is performed by the wireless communication management unit 101 that manages the frequency band and modulation scheme of the wireless access to be used and detects the wireless transmission quality such as RSSI and SINR.

  Since the wireless transmission quality can be grasped by the terminal 100, the terminal is provided with a wireless transmission quality variation determining unit 104 that determines a variation in quality from the wireless transmission quality for a certain period.

  When converting the maximum bandwidth on the terminal 100 side, the maximum bandwidth conversion table 102 is provided, the maximum bandwidth is converted from the use bandwidth, modulation method, etc. managed by the wireless communication management unit 101, and the congestion window control unit 106 is notified.

  When the base station 200 calculates an accurate maximum bandwidth from radio resources that can be allocated to the terminal 100, the base station 200 includes the maximum bandwidth calculation unit 201 to calculate the maximum bandwidth and perform wireless communication of the terminal 100. Notify the management unit 101.

  In the first and second embodiments, the maximum bandwidth is acquired as the wireless access quality. In the third to fourth embodiments, the wireless transmission quality is acquired as the wireless access quality. However, the maximum bandwidth and the wireless transmission quality are acquired as the wireless access quality. Both of them may be acquired. That is, as another embodiment, the combination of the said Embodiment 1 and 3, the combination of Embodiment 1 and 4, the combination of Embodiment 2 and 3, the combination of Embodiment 2 and 4, etc. can be considered.

  Next, (2) the control for adjusting the TCP congestion window size cwnd based on the radio access quality of the present invention will be described. Here, the unit of bandwidth is bps (bit per second), the unit of cwnd and ssthresh is MSS (maximum segment size), and the unit of RTT is second.

In the first and second embodiments in which the maximum bandwidth is acquired as the radio access quality, the congestion window control unit 403 performs one of the following congestion window controls 1, 2, 5, and 6 or a combination of these controls.
[Congestion window control 1] Ssthresh change control based on maximum bandwidth [Congestion window control 2] High-speed cwnd increase control considering maximum bandwidth [Congestion window control 5] cwnd decrease control based on maximum bandwidth [Congestion window control 6] Cwnd increase control based on maximum bandwidth

In the third and fourth embodiments in which the wireless transmission quality is acquired as the wireless access quality, the congestion window control unit 403 performs the next congestion window control 3.
[Congestion window control 3] cwnd control when RTT increases considering radio transmission quality

In the embodiment in which bandwidth control and wireless transmission quality are acquired as wireless access quality, the congestion window control unit 403 obtains the maximum bandwidth acquired in Embodiments 1 and 2, the wireless transmission quality acquired in Embodiment 3, and the RTT. Then, the next congestion window control 4 is performed.
[Congestion window control 4] Change control of cwnd increase / decrease parameter by maximum bandwidth, wireless transmission quality, RTT

  Note that the congestion window control unit 106 of the terminal 100 according to the fifth embodiment also performs any one of the congestion window controls 1 to 6 or any combination thereof from the acquired maximum bandwidth, radio transmission quality, and RTT.

[Congestion window control 1]
The congestion window control 1 is ssthresh change control based on the maximum bandwidth. FIG. 16 shows an operation flow of the congestion window control 1. Here, the maximum bandwidth previously detected by the wireless communication management unit 101 or the access quality monitoring unit 402 is B ₁ , and the newly detected maximum bandwidth is B ₂ .

(1) Step S1
If the maximum bandwidth B ₂ newly detected by the wireless communication management unit 101 or the access quality monitoring unit 402 is different from the previously detected maximum bandwidth B ₁ , the maximum bandwidth is set to the congestion window control unit (106 or 403). Notice. If there is no change in the maximum bandwidth, normal congestion control is performed.

(2) Step S2
If B ₁ <B ₂ , the process proceeds to step S3.

(3) Step S3
Calculating the ssthresh based on RTT obtained from the maximum bandwidth _{B 2} and RTT measurement unit (105 or 401). As a method for determining ssthresh, for example, the following equation can be considered. The following formula shows a value obtained by multiplying cwnd corresponding to the maximum band by γ times in the current RTT. It is conceivable that γ is a constant or varies depending on radio transmission quality and RTT.

  By increasing ssthresh based on the maximum bandwidth, the bandwidth can be used immediately by shifting to the slow start mode.

(4) Step 4
If cwnd ≧ ssthresh, a congestion avoiding mode for controlling to increase / decrease cwnd so as to avoid congestion (step S5), and if cwnd <ssthresh, slow start mode for increasing cwnd twice for each RTT (step S6). Migrate to For the congestion avoidance mode algorithm, for example, an existing congestion avoidance algorithm such as TCP Reno or CUBIC, or the control algorithm of the congestion window control 4, 5, 6 of the present invention may be used.

[Congestion window control 2]
The congestion window control 2 is high-speed cwnd increase control in consideration of the maximum bandwidth. FIG. 17 shows an operation flow of the congestion window control 2. Similar to the congestion window control 1, the congestion window control 2 performs control to increase cwnd at a high speed up to a threshold value based on the maximum bandwidth when the maximum bandwidth increases. However, if the slow start is performed from a state where the cwnd is already large, the cwnd may be suddenly increased and serious congestion may be caused. Therefore, the congestion window control 2 is faster than the normal congestion avoidance mode, and is faster than the slow start mode. Introduces a new high-speed cwnd increase mode that gradually increases cwnd. Here, the normal congestion avoidance mode is any one of those used in existing congestion avoidance algorithms such as TCP Reno and CUBIC. The high-speed cwnd increase mode is obtained by improving the cwnd increase / decrease process with respect to the normal congestion avoidance mode.

  In the congestion window control 2, the value of ssthresh calculated in the congestion window control 1 is set as ssthresh '(step S23). If it is determined in step S26 that cwnd <ssthresh ', the process proceeds to the high-speed cwnd increase mode in step S28. ssthresh 'is set to 0 at high-speed retransmission transfer or time-out, and the high-speed cwnd increase mode is terminated.

  For example, the following control can be considered for the control in the high-speed cwnd increase mode.

(A) From the window size (cwnd ′) when the maximum bandwidth is updated, the following control is performed in which the increase amount is doubled for each RTT with an initial value of 1. cwnd ₀ indicates the window size when receiving the first ACK after changing the maximum bandwidth.

  (B) When the increase amount of cwnd in the normal congestion avoidance mode is larger than the above control, the following control is adopted that employs the larger increase amount.

  FIG. 18 shows the cwnd increase amount in each control mode. FIG. 18 shows the amount of increase in cwnd when the maximum bandwidth is increased at time T.

[Congestion window control 3]
The congestion window control 3 is cwnd control at the time of RTT increase considering radio transmission quality. An operation flow of the congestion window control 3 is shown in FIG. The congestion window control 3 may be employed alone as the congestion control algorithm itself, or may be used in combination with the slow start mode as the congestion avoidance mode. Further, the congestion window control 3 may be implemented as an improvement of an existing congestion avoidance algorithm such as TCP Reno or CUBIC.

(1) Step S31
The RTT measuring unit (105 or 401) measures the RTT for each packet transmission / reception, and notifies the congestion window control unit (106 or 403) as an average value of RTTs for a plurality of times as RTT ₁ .

(2) Step S32
The terminal 100 or the proxy 400 detects a change in wireless transmission quality and notifies the congestion window control unit (106 or 403).

(3) Step S33
An RTT measuring unit (105 or 401) notifies the congestion window control unit (106 or 403) as an RTT ₂ average value of a plurality of RTTs.

(4) Step S34
If the RTT increases (RTT ₂ > RTT ₁ ) before and after the change of the radio transmission quality, the process proceeds to step S35.

(5) Step S35
Assume that the wireless transmission quality is Q, and the higher the Q value, the better the quality. The congestion window control unit (106 or 403) manages the reference value RV _{1 for} determining that the radio transmission quality is low and the reference value RV ₂ for determining that the radio transmission quality is high. If Q <RV ₁ , The increase is determined to be caused by a decrease in wireless transmission quality, and control is performed to increase cwnd in order to maintain the throughput (step S36). If Q> RV ₂ , it is determined that the increase in RTT is caused by congestion, and cwnd is decreased to avoid congestion (S step 38).

As an example of the increase amount of cwnd in step S36, the following equation for maintaining the throughput at RTT ₁ can be considered.

ｃｗｎｄ’：ＲＴＴ増加前のｃｗｎｄ

cwnd ': cwnd before RTT increase

  As an example of reducing cwnd in step S37, control for reducing cwnd as in the following equation is conceivable.

  Further, by using ECN (Explicit Congestion Notification) together, more accurate congestion determination becomes possible.

[Congestion window control 4]
The congestion window control 4 is control for changing the cwnd increase / decrease parameter based on the maximum bandwidth, the radio transmission quality, and the RTT. The operation flow of the congestion window control 4 is shown in FIG. The congestion window control 4 is to change the cwnd increase / decrease parameter according to the maximum bandwidth, the radio transmission quality, and the RTT in the congestion avoidance control process used in arbitrary TCP congestion control such as TCP Reno and CUBIC. General AIMD (Additive-Increase / Multiplicative-Decrease) control of existing TCP congestion control is expressed by the following equation.

  Here, for example, α = 1 / cwnd and β = 1/2 in TCP Reno, and in HighSpeed TCP, control is performed to determine α and β from the current cwnd.

In step S41, S42 in FIG. 20, when different values of the maximum bandwidth _{B 2} newly acquired maximum bandwidth _{B 1} acquired previously in the terminal 100 or the proxy 400, the congestion window controller (106 or 403), the maximum When the bandwidth increases (B ₁ <B ₂ ), the cwnd increase amount during the congestion window size increase processing in TCP congestion control is increased, the cwnd decrease amount during the congestion window size decrease processing is decreased, and the maximum bandwidth decreases. In the case (B ₁ > B ₂ ), control is performed to decrease the cwnd increase amount during the congestion window size increase process and increase the cwnd decrease amount during the congestion window size decrease process. Note that the congestion window size increasing process includes, for example, ACK reception. Further, the congestion window size reduction process includes, for example, packet loss (when duplicate ACK is received or when ACK reception times out).

In step S43 of FIG. 20, similarly to steps S31 to S35 of the congestion window control 3, RTT ₂ newly acquired is larger than RTT ₁ previously acquired by the RTT measurement unit (105 or 401), and the terminal 100 Alternatively, when the wireless transmission quality Q acquired in the proxy 400 exceeds the reference value RV ₂ for determining that the wireless transmission quality managed by the congestion window control unit (106 or 403) is high, it is determined that the increase in RTT is caused by congestion. , Cwnd increase amount is decreased and cwnd decrease amount is increased.

  The above control of the cwnd increase / decrease amount is not limited to AIMD control, and for example, in the following control expression of TCP CUBIC, it is conceivable to perform the same control on parameters C and β.

[Congestion window control 5]
In the congestion window control 5, the current communication throughput (cwnd / RTT) of the terminal 100 is newly set in the terminal 100 or the proxy 400 in the congestion avoidance control process used in arbitrary TCP congestion control such as TCP Reno and CUBIC. When it is larger than the acquired maximum bandwidth B, control is performed to immediately decrease cwnd. The operation flow of the congestion window control 5 is shown in FIG. The amount of decrease in cwnd performed in step S53 in FIG. 22 is controlled to reduce the communication throughput of the terminal 100 to B, which is the maximum bandwidth, for example, as in the following equation.

  The above-described decrease in cwnd is performed in order to prevent congestion from occurring by sending a packet exceeding the maximum bandwidth that the terminal 100 can use.

[Congestion window control 6],
The congestion window control 6 is a congestion avoidance control process used in any TCP congestion control such as TCP Reno or CUBIC, for example, and increases or decreases the amount of cwnd based on the current throughput and maximum bandwidth of communication performed by the terminal 100 and the RTT. It performs control to adjust. As the throughput of communication performed by the terminal 100 is smaller than the maximum bandwidth acquired by the terminal 100 or the proxy 400, the increase amount of cwnd during the congestion window size increasing process in the TCP congestion control is increased, and the congestion window size decreasing process is performed. The decrease amount of cwnd is reduced, and the increase amount of cwnd during the congestion window size increase process is reduced as the current throughput of communication performed by the terminal 100 approaches the notified maximum bandwidth, and the decrease of cwnd during the congestion window size decrease process is reduced. Control to increase the amount. Note that the congestion window size increasing process includes, for example, ACK reception. Further, the congestion window size reduction process includes, for example, packet loss (when duplicate ACK is received or when ACK reception times out). FIG. 23 shows changes in the cwnd increase / decrease amount of the congestion window control 6. The above control can be realized, for example, by calculating α and β in Equation 6 using Equation 9 below. It is conceivable that α ₁ , α ₂ , β ₁ , and β ₂ in Equation 9 are constants, or may be changed depending on radio transmission quality and RTT.

Furthermore, control is performed to increase the increase amount of cwnd as the RTT of communication performed by the terminal 100 increases. In AIMD control such as TCP Reno, a certain amount of cwnd is increased each time an ACK is received. Therefore, the increase in cnwd is faster for communication with a small RTT, and the increase in cwnd is slow for communication with a large RTT. For this reason, although the fairness between communications with different RTTs is not high, the communication throughput difference due to the size of the RTT can be reduced by increasing the amount of increase in cwnd for communications with a larger RTT. This control can be realized, for example, by expressing α _{1 in the} above equation 9 as a function of RTT as in the following equation. It is conceivable that α ₃ is a constant or is varied depending on the radio transmission quality.

The congestion window controls 1 to 6 may be controlled independently or in combination. For example, FIG. 24 shows a control example in which the congestion window control 1 and the congestion window control 6 are combined. In the example of FIG. 24, when the maximum available bandwidth of the terminal 100 is changed from B to B ′, the slow start is performed until cwnd reaches γ (B ′ · RTT) / MSS according to the congestion window control 1. . Thereafter, according to the congestion window control 6, the increase amount is suppressed as cwnd approaches cwnd _{B ′} .

  As mentioned above, although embodiment of this invention was explained in full detail, this invention is not limited to this. For example, the communication format such as LTE described in the above embodiment is an example, and the present invention can be implemented even with other communication formats. The present invention can also be implemented in determining the packet transfer rate of a protocol other than TCP (such as a UDP-based protocol such as QUIC proposed by SCTP or Google).

DESCRIPTION OF SYMBOLS 100 ... Terminal 101 ... Wireless communication management part 102 ... Maximum bandwidth conversion table 103 ... Wireless quality notification part 104 ... Wireless transmission quality fluctuation | variation determination part 105 ... RTT measurement part 106 ... Congestion window control part 200 ... Wireless base station 201 ... Maximum bandwidth calculation Unit 202 ... wireless transmission quality acquisition unit 203 ... wireless transmission quality variation determination unit 300 ... core node 400 ... proxy 401 ... RTT measurement unit 402 ... access quality monitoring unit 403 ... congestion window control unit 500 ... server

Claims (16)

In a packet communication system comprising a radio base station that accommodates a radio terminal and a communication device that forms a transport layer connection of the OSI 7 hierarchical model with the radio terminal,
Radio access quality detection means for detecting radio access quality between the radio terminal and the radio base station;
A packet communication system, comprising: a congestion window control unit that adjusts a congestion window size in the transport layer based on the radio access quality detected by the radio access quality detection unit. The radio access quality detection means calculates a maximum bandwidth that can be used in radio access between the terminal and the radio base station, and if there is a variation in the maximum bandwidth, the congestion bandwidth window The packet communication system according to claim 1, wherein the control means is notified. The packet communication system according to claim 2, wherein the congestion window control means changes a threshold of a congestion window size for shifting from the slow start mode to the congestion avoidance mode based on the notified maximum bandwidth. The packet communication system according to claim 3, wherein the congestion window control means increases the threshold when the notified maximum bandwidth increases. When the notified maximum bandwidth increases, the congestion window control means calculates a second threshold value that is larger than the threshold value of the congestion window size for shifting from the slow start mode to the congestion avoidance mode based on the maximum bandwidth, 3. The packet according to claim 2, wherein in the avoidance mode, the congestion window size is increased more slowly than the normal congestion avoidance mode and more slowly than the slow start mode until the congestion window size exceeds the second threshold. Communications system. In the congestion avoidance control process, the congestion window control means is notified by increasing the amount of increase during the congestion window size increasing process and decreasing the amount of decrease during the congestion window size decreasing process as the notified maximum bandwidth increases. The packet communication system according to claim 2, wherein the amount of increase at the time of congestion window size increase processing is controlled to be small and the amount of decrease at the time of congestion window size decrease processing is largely controlled as the maximum bandwidth decreases. In the congestion avoidance control process, the congestion window control means increases the amount of increase during the congestion window size increase process and decreases during the congestion window size decrease process as the difference between the communication throughput performed by the terminal and the notified maximum bandwidth increases. Decrease the amount, decrease the increase in congestion window size increase processing and increase the decrease amount in congestion window size decrease processing at the time of packet loss as the difference between the communication throughput performed by the terminal and the notified maximum bandwidth difference decreases The packet communication system according to claim 2, wherein the packet communication system is controlled. When the congestion window control means performs control to increase the congestion window size in the congestion avoidance control process, the amount of increase in the congestion window size is increased for a communication with a larger RTT so that a certain amount of throughput is improved in a certain time regardless of the RTT. The packet communication system according to claim 3, wherein the packet communication system is increased. 3. The packet communication according to claim 2, wherein the congestion window control means immediately reduces the congestion window size when the throughput of communication performed by the terminal exceeds the notified maximum bandwidth in the congestion avoidance control process. system. The radio access quality detecting means monitors radio transmission quality in radio access between the terminal and the radio base station, and if there is a change in the radio transmission quality, the radio transmission quality is used as the radio access quality and the congestion window The packet communication system according to claim 1, wherein the control means is notified. RTT measuring means for measuring RTT between the wireless terminal and the communication device,
In the congestion avoidance control process, the congestion window control means determines whether the increase in RTT is due to a decrease in radio transmission quality or due to congestion based on the notified change in radio transmission quality and the measured increase / decrease in RTT. The packet communication system according to claim 10, wherein the congestion window is adjusted based on the determination result. 12. The packet communication according to claim 11, wherein the congestion window control means increases the congestion window size so as to maintain the throughput before the increase in RTT when it is determined that the increase in RTT is due to a decrease in radio transmission quality. system. 12. The packet communication system according to claim 11, wherein when the increase in RTT is determined to be due to congestion, the congestion window control means decreases the congestion window size so as to avoid congestion. If the congestion window control means determines that the increase in RTT is due to congestion, the congestion window control means decreases the increase in the congestion window size during the congestion window size increase process and decreases the decrease in the congestion window size during the congestion window size decrease process. The packet communication system according to claim 11. A congestion control method in a packet communication system comprising: a wireless base station that accommodates wireless terminals; and a communication device that forms a transport layer connection of the OSI 7 layer model between the wireless terminals,
A step in which a radio access quality detection means detects a radio access quality between the radio terminal and the radio base station;
A congestion control method in a packet communication system, characterized in that the congestion window control means includes a step of adjusting a congestion window size in the transport layer based on the radio access quality detected by the radio access quality detection means.   A congestion control program for causing a computer to function as each means according to claim 1.

Images