JPH11261632A - Band allocation control method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To set the allocation priorities of communication bands, in accordance with the kinds and positions job levels, to secure throughput and to shorten response time in important communication and urgent communication by comparing a packet loss event counter with a priority coefficient M and determining the necessity of reduction with respect to a congestion window value and a slow start threshold, which are band control parameters. SOLUTION: A layer 4 processing part 12 of a client side requests connection setting to a server side, in accordance with a communication start request from an application processing part 13. At that time, a packet loss event counter (k) is rest to '0'. It is decided whether preferential band control is to be executed by TCP connection, based on user ID or request application ID informed from higher-order AP, and decides/sets the value of a priority coefficient M. Even in the case of a packet loss event, the transmission control of high priority is executed with respect to pertinent connection. When urgency is not required, M=1 is set.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a bandwidth allocation control method, and more particularly to a bandwidth allocation control method for a protocol layer 4 in a node that executes protocol processing.

[0002]

2. Description of the Related Art At present, the best-effort service is the main service form in the Internet / intranet, and each connection is provided with a band corresponding to a network condition that changes every moment. TCP, which is a typical protocol, detects a change in network conditions based on the detection of a packet loss event or the like, and adjusts the amount of transmission.

For this reason, in addition to a reception notification window (adwnd) indicating the reception capability of the partner node, a congestion window (cwnd) in which the transfer capability of the network is estimated is used to transmit data in a range of min (adwnd, cwnd). (However, min (A, B) indicates that a smaller value of A and B is used). Bandwidth control for adjusting the transmission amount is implemented centering on cwnd, and there is a slow start threshold (ssthresh) in addition to the above parameters, and transmission is performed based on these parameters.

In this type of bandwidth control, there are two phases called a slow start phase and a congestion avoidance phase, and control methods are different. The outline of the method is as follows. The unit of each parameter of the bandwidth control is the number of bytes in the implementation of TCP, but in the following,
Expressed by the number of packets for easy explanation.

FIG. 9 is a flowchart showing a conventional bandwidth allocation control method. First, after the connection is set, the parameters of the bandwidth control are set to ssthresh = adwnd, c
Initially, wnd = 1 is set (step 90). And
In the slow start phase, one data packet (DT packet) is transmitted, and an acknowledgment packet (A
It waits for reception of a CK packet (step 91).

If an ACK packet is received within a predetermined time (step 90: ACK packet reception), c
When wnd <ssthresh (step 92:
YES), cwnd is incremented by 1 (step 93). As a result, next, two packets are transmitted (step 9).
1) Hereafter, D which has been confirmed for every ACK packet received
By the number of T packets, the next transmittable amount, cwnd, is
1, 2, 4, 8, ... and so on.

Also, cwnd is equal to the threshold value ssthres
When h has been reached (step 92: NO), a congestion avoidance phase is entered. In the congestion avoidance phase, cwnd is increased by 1 / cwnd every time an ACK packet is received (step 94), so that the increase is much more gradual than in the slow start phase.

In a typical TCP implementation, when an acknowledgment is not received and time monitoring times out (step 9).
1: When an acknowledgment timeout occurs, or when a certain number (usually three) or more duplicate ACKs are received (Step 91: Duplicate A)
CK reception), it is determined that a packet loss has occurred, that is, a DT packet loss event. Note that the duplicate ACK reception means that a plurality of ACK packets having the same reception sequence number are continuously received.

When it is determined that the DT packet is lost, the parameters of the band control are adjusted. In particular, in the case of duplicate ACK detection, ssthresh is set to min (cwnd, adwn
d) / 2, where cwnd is new cwnd = old cwnd / 2
(Step 95). Also, in case of timeout,
ssthresh to min (cwnd, adwnd) /
2 and cwnd is set to new cwnd = 1 (step 9
6) (for example, WR Stevens, TCPIl)
trustedVol. 1 Chapter21,
See Addison Wesley, 1994).

On the other hand, a selective response (SACK) function aimed at speeding up error recovery has been introduced as an Internet document in 1996 as an option of TCP.
Specified by FC2018. According to this, in the SACK information, for all the packets that have been received discontinuously (that is, due to packet loss) on the receiving side, the section in which the reception was successful is added to the ACK packet and specifically notified.

The transmitting side uses this information to retransmit only the lost packet. The relationship with bandwidth control (congestion control)
Since the existing functions should be preserved in RFC 2018, when a packet loss is detected based on the SACK information, the same band control parameter as that at the time of detecting the duplicate ACK is adjusted. As described above, when a packet loss event is detected based on duplicate ACK, acknowledgment timeout, and SACK information, control to uniformly reduce the transmission amount is performed regardless of the user level or application type using the connection. Has become

[0012]

With the diversification of information communication services and network environments, service grades are also required for best-effort services. For example, even when the network is congested, high priority such as important / emergency is required. There is a request to secure a band in a communication connection. However, in such a conventional bandwidth allocation control method, for example, in the current TCP connection, control is performed to uniformly reduce the communication bandwidth when a packet loss event is detected, regardless of the communication requirements of the application.

For this reason, when the network is congested, there is a problem that it is not possible to secure a band in connection of high priority communication such as important / urgent. For example,
In intranets supposed to be used internally, it is conceivable to assign priorities to the allocation of communication bandwidth according to the type of business and job position to secure throughput and reduce response time in important communications and emergency communications. However, there is a problem that it cannot be realized by the bandwidth control.

The present invention is intended to solve such a problem. Even if a packet loss event is detected, the transmission amount is not immediately reduced, but the TC of an important communication or an emergency communication is reduced.
An object of the present invention is to provide a bandwidth allocation control method capable of securing a throughput and shortening a response time for a predetermined connection by performing a bandwidth control operation for securing a communication bandwidth in the case of a P connection. I have.

[0015]

In order to achieve the above object, a band allocation control method according to the present invention provides a function of measuring a packet loss event, a packet loss event counter k holding the number of lost packets, and a packet lost event counter k. A priority coefficient M for determining whether or not to reduce the value of the band control parameter at the time of an event is provided.
Is compared with the priority coefficient M to determine whether it is necessary to reduce the congestion window value (cwnd) and the slow start threshold value (ssthresh), which are bandwidth control parameters.

In this case, the priority coefficient M is specified by the host application at the time of a connection setting request, or
It is set in advance in the CP module. When setting a connection, the packet loss event counter k is set to 0.

Therefore, when a packet loss event is detected, 1 is added to the packet loss event counter k, and it is checked whether or not k has reached the priority coefficient M. Here, if k <M, ssthresh and cwnd are not reduced, and only retransmission of lost packets is performed thereafter.

When k ≧ M, sslhres
h and cwnd are reduced, lost packets are retransmitted, and k = 0 is reset. The initial value of the priority coefficient M is 1, in which case the conventional control operation without priority control is performed, and the band control parameters ss1hresh and cwnd are immediately reduced.

[0019]

Next, the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing a communication system to which a band allocation control method according to an embodiment of the present invention is applied. In FIG. 1, an end node 1 is connected to an information communication network 2 and communicates with a partner end node 1.

The end node 1 includes a lower layer processing unit 1
1, layer 4 processing unit 12, application processing unit 13
It is composed of The lower layer processing unit 11 performs processing below the protocol layer 3, that is, layer 1 such as electrical matching with a communication line, layer 2 such as frame assembly / disassembly, and layer 3 (such as I / O) in routing.
P).

The layer 4 processing unit 12 performs processing of data transmission and reception based on release of connection setting of layer 4 (here, TCP), flow control, and the like. The information communication network 2 includes a communication line and a relay node. As parameters of the bandwidth control of the protocol layer 4 (TCP), a reception notification window (adwnd), a congestion window (cwnd), a slow start threshold (ssth)
refresh).

In the present invention, in addition to these parameters,
A variable k for counting the number of lost packets and holding the number of lost packets, and a priority coefficient M for determining whether to reduce the bandwidth control parameter when detecting a lost packet event are provided. The number of packets that can be sent is min
(Cwnd, adwnd), which changes during communication with the change of each parameter (however, min (A,
B) indicates that the smaller value of A and B is used).

Next, with reference to FIG. 2, a description will be given of a data transmission processing operation focusing on band control as an operation of the present invention. FIG. 2 is a sequence diagram showing the operation of the present invention. First, the layer 4 processing unit 12 (T
The CP) makes a connection setting request to the server in response to a communication start request from the application processing unit 13 (upper AP). At this time, the packet loss event counter k is reset (cleared) to 0.

The higher-level AP on the server side then determines whether or not to perform preferential bandwidth control over the TCP connection based on the user ID or the requested application ID notified from the higher-level AP on the client side, and determines a priority factor. M
Determine and set the value of. For example, in the case of urgent communication, a value larger than 1 is set to M according to the degree of urgency.

Thus, even in the case of a packet loss event, high-priority transmission control is performed on the connection. Actually, in the case of a packet loss event, the transmission coefficient is not immediately reduced, and the priority coefficient M is compared with the packet loss event counter k. Only when k ≧ M, that is, the packet loss event occurs M times Each time, the transmission amount is reduced.

When ordinary communication is sufficient without urgency, M = 1 is set. The initial value (default value) of the priority coefficient M is set to 1, and the operation of the conventional control without the priority control is performed. Thus, in the event of a packet loss event,
With each occurrence, the transmission amount is immediately reduced.

FIG. 3 is an explanatory diagram showing a band change by band allocation control. When the connections A to C are set via the network, in the conventional bandwidth allocation control, as shown in FIG. 3A, the bandwidth control parameters ssthresh and cw are immediately set in response to the occurrence of the packet loss event.
nd is reduced, and the bandwidth allocation of all the connections A to C is reduced.

On the other hand, according to the present invention, FIG.
As shown in (b), in the packet loss event, the transmission amount of the normal communication (M = 1) connections B and C is reduced every time the packet loss event occurs. Is transmitted M times, the transmission amount is reduced. Therefore, high priority connection A
Is smaller than the other connections B and C, and the bandwidth allocation is prioritized.

In this way, even if a packet loss event or the like is detected, the transmission amount of the high-priority connection is not reduced immediately, but is reduced every M detections. When the packet loss / delay is detected, the transmission amount is reduced uniformly. As a result, the bandwidth released by the conventional host / terminal is acquired and used by the host / terminal of the present invention.

Therefore, TCP for important communication and emergency communication
In the case of a connection, by performing a band control operation to secure a communication band, it is possible to secure a throughput and shorten a response time. In addition, since the mechanism is realized on the information transmitting side, the server side can set the priority coefficient M according to the user level of the connected client or the application type, thereby enabling unlimited use of high priority. Can be avoided.

FIG. 4 is an explanatory diagram showing a change in the transmission amount.
Here, the change of cwnd is shown by taking the case where the priority coefficient M = 2 as an example. For example, when the transmission is started at time T0, the cwn of the conventional method is performed by the slow start phase.
Both d41 and cwnd42 of the present invention increase by the number of acknowledgment packets.

Here, when a packet loss event occurs at time T1, the transmission amount is immediately reduced in the connection of the conventional system. On the other hand, in the connection of the present invention, since M = 2, the transmission amount is not reduced at time T1 because it is the first packet loss event, and is reduced for the first time at the second packet loss event at time T2. Therefore, the cwnd4 of the present invention is better than the cwnd41 of the conventional method.
2, the priority becomes higher in the bandwidth allocation control.

[0033]

Next, a first embodiment of the present invention will be described with reference to FIG. FIG. 5 is a flowchart showing a processing procedure of the bandwidth allocation control method according to the first embodiment of the present invention. First, after the connection is set, the parameters of the bandwidth control are initialized to ssthresh = adwnd, cwnd = 1 (step 50), the priority coefficient M of the connection is set, and the packet loss event counter k is initialized (step 50). Step 51).

Then, a slow start phase is entered,
One data packet (DT packet) is transmitted,
It waits for the receipt of an acknowledgment packet (ACK packet) (step 52).

If an ACK packet is received within a predetermined time (step 52: ACK packet reception), c
When wnd <ssthresh (step 53:
YES), cwnd is incremented by 1 (step 54). As a result, next, two packets are transmitted (step 5).
2) Hereafter, D which has been confirmed for every ACK packet received
By the number of T packets, the next transmittable amount, cwnd, is
1, 2, 4, 8, ... and so on.

Also, cwnd is equal to the threshold value ssthres
When h has been reached (step 53: NO), a congestion avoidance phase is entered. In the congestion avoidance phase, cwnd is increased by 1 / cwnd every time an ACK packet is received (step 55), so that the increase is much slower than in the slow start phase.

In a typical TCP implementation, when an acknowledgment is not received and time monitoring times out, or when a certain number (usually 3) or more duplicate ACKs are received (step 5).
2: Duplicate ACK reception / acknowledgement timeout), it is determined that packet loss has occurred, that is, a DT packet loss event. Note that the duplicate ACK reception means that a plurality of ACK packets having the same reception sequence number are continuously received.

When it is determined that the DT packet is lost, the parameters of the band control are adjusted as follows. First, the packet loss event counter k is incremented by 1 (step 56).
Is compared with the priority coefficient M (step 57). Here, if k <M (step 57: NO), ssthr
The process returns to step 52 without reducing esh and cwnd, and only retransmits the lost packet.

If k ≧ M (step 57: YE
S), set cwnd to min (old cwnd, adwnd) /
2, and ssthresh is set to the new cwnd (step 58). Then, after the packet lost event counter k is initialized (step 59), the process returns to step 52 to retransmit the lost packet.

Next, a second embodiment of the present invention will be described with reference to FIG. FIG. 6 is a flowchart showing a processing procedure of the bandwidth allocation control method according to the second embodiment of the present invention. Here, as compared with the first embodiment, when k <M in step 57, the process does not immediately return to step 52, and in the case of the slow start phase, ssthre
sh is reset.

That is, cwnd is compared with ssthresh (step 60), and cwnd <ssthresh
In the case of h (step 60: YES), it is determined that it is the slow start phase, and sstresh is set to c.
wnd is set (step 61), and the process returns to step 52. If cwnd ≧ ssthresh (step 60: NO), it is determined that it is not the slow start phase, and the process returns to step 52 as it is.

As described above, in the slow start phase, cwnd increases exponentially, so if the network congestion situation is very bad, continuing the slow start phase as it is will further worsen the congestion. In the present embodiment, in consideration of application to a network environment in which a very bad congestion situation is likely to occur, when k <M, cwnd is set as ssthresh only in the slow start phase. So that we can immediately move to the congestion avoidance phase,
It is possible to maintain a high-priority connection throughput without aggravating congestion.

Next, a third embodiment of the present invention will be described with reference to FIG. FIG. 7 is a flowchart showing a processing procedure of the bandwidth allocation control method according to the third embodiment of the present invention. Here, compared to the first embodiment, in the case of an acknowledgment timeout among the packet loss events, the process proceeds to step 58, and ssthresh and cwnd are unconditionally reduced. .

The time-out occurs when the network congestion is generally very bad. Specifically, the transmission-side time-out occurs on the transmission side due to the loss of a plurality of data packets and the loss of the response packet. Become. On the other hand, the meaning of the timeout is slightly different depending on the operation characteristics of the relay node (router) constituting the network.

RED (random early) in which multiple packets are not easily discarded in one round trip
In an environment using a router that performs transmission buffer management of the type (Detection), the occurrence of a timeout indicates that the congestion situation is very bad. On the other hand, a simple FI
In a router that performs transmission buffer queue management by the FO method, a plurality of packets are more likely to be discarded than in a RED type router.

Therefore, as in the present embodiment, RE
In a network environment using a D-type router, more appropriate bandwidth control can be realized by dividing the bandwidth control processing for a lost event into duplicate ACKs and timeouts. That is, since the timeout in such a case indicates that the network congestion situation is very bad, even if the connection has a high priority, cw is temporarily set.
By reducing nd, it is possible to avoid a deteriorating network congestion state, which can be used to smoothly perform priority bandwidth allocation thereafter.

Next, a fourth embodiment of the present invention will be described with reference to FIG. FIG. 8 is a flowchart showing a processing procedure of the bandwidth allocation control method according to the fourth embodiment of the present invention. Here, the second embodiment and the third embodiment are combined.

That is, if k <M in step 57, the process does not immediately return to step 52. In the case of the slow start phase, ssthresh is reset, and in the case of an acknowledgment timeout, step 58 is performed.
Then, ssthresh and cwnd are unconditionally reduced.

This embodiment is directed to a case where the traffic situation and the characteristics of the network environment described in the second and third embodiments are combined, and a RED type router is used. It is assumed to be applied to a situation where congestion is bad in a congested network environment. As a feature, it is possible to maintain high-priority connection throughput without further deteriorating congestion, and when a timeout occurs, recover the deteriorating network congestion situation and smoothly perform priority bandwidth allocation thereafter. Can help you do that.

In the above description, the operation when the SACK (selective response) function is supported as a method of detecting a packet loss event is almost the same as that at the time of receiving the duplicate ACK, but the packet loss event counter k Is not the added value to +1 but the lost number found from the SACK information (step 56).

The band control parameter ssthres
The reduction of h and cwnd is reduced by the same formula as in the conventional method. That is, ssthresh = min (cw
nd, adwnd) / 2, new cwnd = old cwnd / 2
And

[0052]

As described above, according to the present invention, when the network is congested, the bandwidth of the connection of the high priority communication such as important / emergency communication is reduced so that the high throughput / low delay compared with other connections can be avoided. Can be realized. In particular, in-house network / intranet position /
It is easy to realize a service grade according to a department / service content.

This is a technology capable of coping with diversification of QoS (service quality) in a best effort type service. Furthermore, a priority coefficient can be used as a management / control means for band allocation, which is one of network management. Accordingly, only the end device that wants to set the M value according to the priority can cope with it, and it is not necessary to particularly deal with the relay node / router, so that the introduction is easy.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a communication system to which a bandwidth allocation control method according to an embodiment of the present invention is applied.

FIG. 2 is a sequence diagram showing an operation of the present invention.

FIG. 3 is an explanatory diagram showing a band change by band allocation control.

FIG. 4 is an explanatory diagram showing a change in a transmission amount.

FIG. 5 is a flowchart illustrating a processing procedure of a bandwidth allocation control method according to the first embodiment of the present invention.

FIG. 6 is a flowchart illustrating a processing procedure of a bandwidth allocation control method according to a second embodiment of the present invention.

FIG. 7 is a flowchart illustrating a processing procedure of a bandwidth allocation control method according to a third embodiment of the present invention.

FIG. 8 is a flowchart showing a processing procedure of a bandwidth allocation control method according to a fourth embodiment of the present invention.

FIG. 9 is a flowchart showing a conventional bandwidth allocation control method.

[Explanation of symbols]

1 end node, 2 information communication network, 11 lower layer processing unit, 12 layer 4 processing unit, 13 application processing unit.

Claims (3)

[Claims] 1. A packet loss event counter in a communication protocol for transmitting a packet within a smaller value between a reception notification window value indicating a reception capability of a partner node and a congestion window value estimating a transfer capability of a network. A value and a predetermined priority coefficient are provided. When a packet loss event is detected, the packet loss event counter value is incremented by 1 and then compared with the priority coefficient. If the packet loss event counter value is smaller than the priority coefficient, congestion occurs. A bandwidth allocation control method characterized by reducing a congestion window value when a packet loss event counter value reaches a priority coefficient without changing a window value. 2. A packet is transmitted within a range of a smaller value between a reception notification window value indicating a reception capability of a partner node and a congestion window value estimating a transfer capability of a network, and an acknowledgment is not received. Duplicate AC that receives a predetermined number of ACK packets with the same reception sequence number when time monitoring times out or
In a communication protocol for detecting occurrence of a packet loss event when K reception is performed, a packet loss event counter value and a priority coefficient are provided, and when a packet loss event is detected due to timeout, the congestion window value is immediately reduced, The packet loss event counter value is reset to 0, and upon detection of a packet loss event due to duplicate ACK reception, the packet loss event counter value is added by one and then compared with the priority coefficient, and the packet loss event counter value is smaller than the priority coefficient A bandwidth allocation control method characterized in that the congestion window value is not reduced in such a case, and the congestion window value is reduced when the packet loss event counter value reaches a priority coefficient. 3. The bandwidth allocation control method according to claim 1, further comprising a slow start phase and a congestion avoidance phase as a congestion window value management phase, wherein a packet loss event is detected in the slow start phase. And immediately shifting to a congestion avoidance phase irrespective of a packet loss event counter value.