KR100798920B1 - Method for Congestion Control of VoIP Network Systems Using Extended RED Algorithm and Apparatus for thereof - Google Patents

Abstract

The present invention manages the TCP / UDP packet drop rate in a node by extending the RED management algorithm in a general Internet or differential service network, thereby interworking with a PLC (Packet Loss Concealment) function for VoIP service to improve VoIP call performance. The present invention relates to a queue management apparatus that enables congestion control at a node. In a packet network for providing an Internet service, there is no in-network congestion due to traffic fluctuations in order to prevent network congestion at each node and to eliminate the congestion in advance. In order to improve the performance of packet loss compensation for end-to-end Internet services, especially VoIP calls, efficient queue management is performed for inter-node packet transmissions in intermediate nodes.

VoIP Gateway, Packet Loss Concealment, Random Early Detection (RED), Voice Quality

Description

Congestion control method and apparatus therefor in a BIPIP network which extends RED method

1 is a configuration diagram of an embodiment of a network for a VoIP service to which the present invention is applied.

2 is a configuration and block diagram of a gateway or node system to which the present invention is applied.

3 is a packet drop rate graph for congestion avoidance processing parameters for congestion of traffic when linked with packet drop control and packet loss concealment of a router queue management or VoIP coder or gateway to which the present invention is applied.

4 is a flowchart of a marking and shaping process related to packet classification and silence according to the present invention.

5 is a flowchart of interworking between the VoIP control management to which the present invention is applied and the queue control management and the PLC helper message using the extended RED mechanism.

In the packet network for providing an Internet service, the present invention provides a packet loss for an end-to-end Internet service, especially a VoIP call, in which there is no in-network congestion due to traffic fluctuations in order to prevent network congestion at each node and eliminate it in advance. A method and apparatus for controlling congestion in a VoIP network using an extended RED execution mechanism that performs efficient queue management for inter-node packet transmission in an intermediate node to improve compensation performance.

As more packets exist on the network, the performance of the network naturally decreases. This rapid deterioration of network performance is called congestion. The biggest cause of congestion is congestion. As an algorithm for avoiding congestion on the network, a tail drop method was initially used.

The tail drop algorithm is a technology for managing the length of the router queue. When the maximum length is set for each buffer queue of a router, which is one of the network nodes, and when the length is exceeded, the length of the buffer queue decreases below the set value. Continue to discard input packets. However, the tail drop algorithm not only has a buffer overflow and a global synchronization problem, but also has a problem in that performance decreases rapidly when the network is expanded at high speed and large capacity.

In order to solve the above problems, a random early detection (hereinafter referred to as 'RED') algorithm has been proposed. The RED algorithm attempts to keep the average queue size low in conjunction with the end-to-end congestion control algorithm. It monitors the traffic load and drops packets with stochastic techniques when the congestion begins to increase. Detecting traffic and slowing down transmission is the main focus. Therefore, this RED algorithm is designed to be applied to the operation of TCP mainly in IP network environment.

However, the RED algorithm is applying a uniform packet drop probability, resulting in a drop in network efficiency due to a low throughput problem due to high continuous drop. In addition, since various differential service (DiffServ) providing networks to be developed in the future are not sufficient quality performance for real-time services such as VoIP, such a conventional RED alone is difficult to expect a smooth service.

The present invention solves the above problems and improves the quality of voice calls on the Internet, thereby improving the packet processing throughput so that the VoIP service is robust against the congestion of traffic in the network and the optimal efficient network operation without idle network resources. It is an object of the present invention to provide a method for further reducing packet loss rate.

In addition, the present invention provides a method for controlling TCP and UDP packet loss at a receiving gateway or a terminal neighbor node, and a method for controlling transmission on a bottleneck link based thereon, and a queue management algorithm that minimizes voice packet loss. For the purpose of

The congestion control device of the VoIP network for achieving the above object of the present invention is a queuing module for managing the queue control for packets received through the VoIP network, a packet scheduler for transmitting the packets processed in the queuing module to the next node And a scheduler module including a PLC helper for creating and sending an RTCP message associated with PLC processing information to an upstream node and receiving an analysis of traffic or packet control messages input from the outside and analyzing the received traffic or packet control messages of the queuing module and the scheduler module. A device comprising a VoIP management module for managing a PLC helper,

In particular, the traffic or packet control related message received by the VoIP management module is one or more messages selected from a message transmitted from an operator, a message transmitted from a VoIP support server or a network management system, or a message transmitted from a called terminal. ,

The PLC helper is based on the input message from the VoIP management module, loss correlation index, round trip time (RTT), loss in PLC circuit of the corresponding codec based on UDP / TCP packet loss rate collected by the VoIP packet receiving end gateway. An RTCP message associated with the PLC processing information, including processing capability information and jitter values for prediction of congestion in the network, is created and sent to the upstream node of the ongoing VoIP session every RTCP cycle.

In this case, before the queuing module manages the queue control for the packets received through the VoIP network, a classification and demultiplexing module for demultiplexing the received packets by differential services and classifying and demultiplexing It is preferable to further include a marking / safety module for processing marking or shaping of each packet processed by the module.

In this case, if the traffic packet input from the external network corresponds to a TCP packet, the classification and demultiplexing module performs demultiplexing with a TCP source, and if it does not correspond to a TCP packet, then a non-TP source. Demultiplexing can be performed with (non-TCP source).

In addition, the marking / safety module, for the demultiplexed packet, if the packet is the best packet corresponding to the best service of the IETF differential service, marking the service level agreement (SLA) for each packet in the best service processor. Or a shaping process, and when the packet corresponds to a band-allocated packet of an Expedited Forwarding (EF) band guarantee service, a marking or shaping process corresponding to a service level agreement (SLA) for each packet is performed by a premium machine. Alternatively, when the packet is a packet corresponding to an AF (Assured Forwarding) service of a differential service, it is preferable to perform marking or shaping processing corresponding to a service level agreement (SLA) for each packet in the AF transmission band allocator.

In this case, the marking / shaping module performs the demultiplexing process on the demultiplexed packet through the BTP source device of the classification and demultiplexing module. In the case of the silent packet, if the set PLC upper limit value corresponding to the loss coping force of the codec exceeds the threshold value, the AF transmission band allocator performs the shaping process, and when the VoIP voice packet is the silent packet, If it does not exceed the threshold value, the AF transmission band allocator may mark the AF packet.

At this time, the queuing module stores the AF packet inputted from the marking / shaping module to perform queue control management according to the extended RED algorithm according to the present invention, and from the best service processor of the marking / shaping module. It is preferable to include a FIFO queue that stores the best packet of the input and performs the corresponding queue management algorithm.

In particular, the RED queue of the queuing module has a first queue management algorithm that performs packet drop based on an average queue length estimation according to a parameter transmitted from the VoIP management module, and a terminal side in consideration of a limitation of packet loss compensation in a VoIP service. It is preferable to include a traffic / label controller for managing queue control by selectively using a second queue management algorithm that performs packet drop using VoIP packet processing information.

In this case, the congestion control apparatus in the network according to the present invention may be implemented by being included in a router, a router switch, a logic circuit and a processor having a digital signal processor (DSP) structure.

In addition, the congestion control method of the VoIP network for achieving the above object of the present invention,

Queuing for packets received via the VoIP network;

A packet scheduling step of transmitting the queue control managed packet to a next node; And,

The VoIP management module receives and analyzes a traffic or packet control related message, which is one or more messages selected from a message transmitted from an operator, a message transmitted from a VoIP support server or a network management system, or a message transmitted from a called terminal. VoIP management step for managing the queue control step and packet scheduling step,

In particular, the VoIP management step is based on the UDP / TCP packet loss rate collected by the VoIP packet receiving end gateway from the PLC helper when the message is a PLC helper drive request message as a result of analyzing the traffic or packet control related message. Upgrading an ongoing VoIP session by creating an RTCP message associated with the PLC processing information, including loss correlation index, round trip time (RTT), loss processing capability information in the PLC circuit of the corresponding codec, and jitter for prediction of congestion in the network. Send it out every RTCP cycle to the node,

The queuing may include a first queue management algorithm that performs packet drop based on an average queue length estimation for an externally transmitted VoIP packet, or a packet using terminal-side VoIP packet processing information in consideration of a limitation of packet loss compensation in a VoIP service. A queue control is managed by selectively using a second queue management algorithm that makes a drop.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. In addition, the same reference numerals are used for parts having similar functions and functions throughout the drawings.

1 is a diagram schematically illustrating a process of providing a VoIP service on a network including a congestion control device according to an embodiment of the present invention. The connection between the router, the router switch, and the IP / ATM network in the provider-provided internal network on the network is the same as that of the conventional network configuration and thus is omitted.

Referring to FIG. 1, a call of a voice call customer 121 in the premises network 120 is transmitted over a private leased line (PBX) 122 and a small gateway 123 via an Internet dedicated line. At this time, the call according to the H.323 protocol (GK) 102, the call according to the SIP protocol (Proxy) 102 through the provider VoIP base network 100 through the destination side gate keeper / Through the proxy 101 and the gateway 131, the call is transmitted to the subscriber in the public telephone network 130 or the terminal connected to the mobile network 110. The call to the public Internet 140 is a network structure in which VoIP calls are performed through connection to an Internet phone or connection with another network.

In this case, the VoIP service includes a network structure capable of guaranteeing QoS by connecting an access network having an access gateway 105 accommodating various multimedia services and a backbone network (core) 100 according to the configuration of an operator network. In addition, various protocols (H.323, SIP, or MGCP) coexist at present, but include soft switches 103 and 104 in an evolving network structure capable of integrated processing.

In FIG. 1, the network structure may include an IP-PBX capable of providing an IP-based service for VoIP, a USB-phone having an xDSL connection, a soft phone, a video IP phone, an Internet phone, a DMB terminal, and the like. .

Hereinafter, a congestion control method and a device therefor in a network according to the present invention will be described in detail with reference to FIGS. 2 to 5.

2 is a view showing a schematic structure of a congestion control apparatus on a network according to an embodiment of the present invention.

Referring to FIG. 2, the congestion control apparatus in the network according to the present invention includes a classifier / demultiplexer 200, a marking safer module 230, and a queuing module. 240, a scheduler module 250, and a VoIP manager module 290.

The classifying and demultiplexing module 200 receives a traffic packet 201 input from an external network, and then performs demultiplexing by dividing the packet for each differential service. Preferably, when the packet corresponds to a TCP packet, the demultiplexing process is performed by a TCP source 210, and when the packet does not correspond to a TCP packet as in the case of UDP corresponding to VoIP, The demultiplexing process can be performed by the non-TCP source 220.

In this demultiplexed packet, the marking / shaping process for congestion control is performed in the marking / shaping module 230.

The marking / shaping module 230 receives a packet transmitted from the classification and demultiplexing module 200 and then performs marking and shaping processing according to subscriber service and band allocation according to the user profile of the packet. To this end, the marking shaping module 230 includes a best service processor (BE) 234, a premium 234, and an AF transmission band allocator 232. It is.

For the packet demultiplexed with the TCP source 210, if the packet is a best packet corresponding to the best service of the IETF Differentiated Service, the best service processor (BE) 234 A marking or shaping process corresponding to a packet-level service level agreement (SLA) is performed at. If the packet corresponds to a band-allocated packet of an Expedited Forwarding (EF) band guarantee service, the marking unit 238 performs marking or shaping processing corresponding to a service level agreement (SLA) for each packet. If the packet corresponds to an Assured Forwarding (AF) service of a differential service (Diffsrv), a marking or shaping process corresponding to a service level agreement (SLA) is performed for each packet in the AF transmission band allocator 232. Is performed.

Packets demultiplexed through a non-TCP source 220, such as VoIP voice packets, are marked at the premium 238 if the packet is a non-mute packet. However, when the set VoIP upper limit value corresponding to the loss coping ability of the codec exceeds the threshold value when the VoIP voice packet is a silence packet, since a specific silence for performance other than voice is input, congestion is expected. The AF transmission band allocator 232 performs the safe on the AF packet. If the VoIP voice packet is a silent packet and the upper limit of the PLC does not exceed the threshold, the AF transmission band allocator 232 performs marking on the AF packet. In this case, the marking may use an IETF Differentiated Service Code Point (DSCP) code point or type of service (ToS) bit of a packet header.

In this way, the marking / shaping module 230 inputs the respective packets, which have been marked or shaved, into the queuing module 240.

The queuing module 240 not only performs queue control management on the input packet, but also links with a traffic / label controller 242 including label control in multi-protocol label switching (MPLS). This is done. The queuing module 240 includes a RED queue 244 and a FIFO queue 248 for queue control management of input packets. At this time, the RED queue 244 stores the AF packet input from the marking / shaping module 230 to perform queue control management according to the extended RED algorithm according to the present invention, the FIFO queue 248 is the Queuing and managing according to the corresponding queue management algorithm is performed by storing the best packet inputted from the best service processor (BE) 238 of the marking / shaping module 230.

In this case, when the packet is a premium packet input from the premium unit 234 of the marking / shaping module 230, a packet scheduler (Packet Scheduler) is not performed for the AF packet or the best packet. 252). The premium packet processed first by the packet scheduler 252 is transmitted to a downstream node or a corresponding terminal according to the scheduling scheme (256).

The packet processed by the queuing module 240 is transmitted to the scheduler module 250.

The scheduler module 250 transmits a packet stored in the queuing module 240 to a next node according to a scheduling algorithm promised by the system. The scheduler module 250 includes a packet scheduler 252 and a PLC helper 258. The packet scheduler 252 transmits a packet transmitted from the queuing module 240 to a downstream node or a corresponding terminal according to a scheduling scheme (256), and the PLC helper 258 is an upstream node of a VoIP session. (Upstream node) 258 generates and transmits a control message for PLC-related information constructed based on the input message 280 from the VoIP management module 290 at every RTCP cycle.

The VoIP management module 290 receives an operator message or a traffic message from a VoIP support server or a network management system or a packet control related signal message or a traffic control message at a called terminal (260), and then analyzes the message. The queuing module 240 or the PLC helper 258 is managed in response to the analysis result. That is, when the message received as a result of the message analysis is a PLC helper message to be requested to the PLC helper, the PLC helper 258 manages VoIP at every RTCP cycle to an upstream node 258 of the VoIP session. Based on the input message 280 from the module 290 to process the PLC helper function to generate and transmit a control message for the PLC-related information. If the message received as a result of the message analysis is not a PLC helper message, the queueing module 240 requests a queue management parameter and exchanges a message 270 for confirming adaptation. On the other hand, if the queue management parameter requested from the VoIP management module 290 is a parameter to be performed in its own node, the queuing module 240 performs queue management using the extended RED algorithm according to the present invention. If the parameter is updated, queue control management is performed by updating the RED parameter value.

In this case, the method for managing a queue using the extended RED algorithm according to the present invention performed by the queuing module 240 is an algorithm characterized in that to manage the RED queue in conjunction with a PLC function, a low pass in the communication network In estimating the average queue length of the exponentially weighted moving average of the RED method using a lowpass filter, packet drop is handled differently according to the range based on the queue length, thereby improving output and end-to-end yield and reducing packet loss. Is the transmission method.

Hereinafter, the extended RED algorithm according to another embodiment of the present invention will be described in detail with reference to FIG. 3 and equations that can be used in the preferred extended RED algorithm applicable to the present invention.

First, using the following equations (1) and 2 will be described packet drop probability that can be applied to the extended RED algorithm according to another embodiment of the present invention.

Equation 1 is for estimating an average queue length, and is an equation for estimating an average queue length in preparation for congestion. Here, S _n is the average queue length, w is the weight applied to the current queue length, q _n is the current queue length, m is the system input load throughput value.

Equation 2 shows a packet drop probability value Pd applied to the extended RED algorithm of the present invention. In particular, the AF transmission band allocator of the marking and shaping module 230 may be used. The packet drop probability value Pd performed by the traffic label controller 242 in the RED queue 244 that performs queue control management on the packet input from 232. Where k _l is the lower average queue length, k _h is the upper average queue length, and max _p is the maximum drop probability value between k _l and k _h .

If the length of the queue estimated using Equation 1 is less than the lower average queue length k _l , packet drop is not performed. If the length of the estimated queue is between two values of a lower average queue length (k _l ) and an upper average queue length (k _h ), or the length of the estimated queue is equal to the upper average queue length (k _h ). If the value falls between two values of the maximum buffer size (K), packet drop is performed by applying each corresponding packet drop probability value defined in Equation (2).

3 is a packet drop rate graph illustrating a relationship between parameters of a packet drop probability applied to an extended RED algorithm according to another embodiment of the present invention. In this case, as described above, max _p in Equation 2 means a maximum drop probability value between k _l and k _h values, and therefore, the average queue estimated by Equation 1 in the extended RED algorithm of the present invention. A packet drop may be performed by taking a linear proportional value from 0 to max _p according to the average queue length in a section including a length.

In this case, the above-described parameter values are inputted and operated through the VoIP management module 290 by the operator or the corresponding request message 260. In this case, when the message is a PLC helper drive request message, the VoIP management module 290 transmits the message to the PLC helper 258 (280), and the PLC helper generates an RTCP-XR packet to generate a voice signal for each period. Loss Correlation Index (ρ _UT ) based on UDP / TCP packet loss rate collected by the receiving end-to-end VoIP gateway, Round Trip Time, Loss processing capability information in PLC circuit of corresponding codec, Jitter for prediction of congestion in network The value is sent via higher node match 258. This traffic information is collected to VoIP server or gateway system through RTCP message between nodes to estimate network condition. If the data collected by the VoIP server or the gateway system in the above manner exceeds a certain value, it is determined that the network is congested. The received information is transferred to the VoIP management module 290, and then the corresponding information is provided to the traffic label controller 242 of the queuing module 240 by the above-described procedure, and the controller controls the packet drop for the RED queue. Drop management is performed according to Equation 3 below.

는 VoIP 코덱의 PLC회로에서의 음성호에 대한 요구 품질을 위해 인접한 몇 개까지의 패깃 손실을 보상할 수 있는지에 대한 패킷 손실률로 표시된 처리 능력에 해당하는 값이다. 음성신호에 대해서도 단말 또는 게이트웨이 간 RTT(Round Trip Time)의 평균 및 해당 세션에 대해 산출하여, 각각 상기 수학식 3의 RTT 및 I 값으로 적용한다. i 및 j값은 해당 세션 번호 및 평균을 구하는데 쓰이는 시간 구간 값이고, 본 명세에서 사용되는 파라미터는 포트당 클래스당 산출하고, max_p는 클래스당 적용한다. 이때

연산은 평균 RTT가 개별 값보다 큰 경우 플러스 연산을 하고, 그렇지 않은 경우 마이너스 연산을 취한다. In Equation 3, N is the number of input packets since the last drop in the node performing RED queue management. However, in the case of VoIP packet shaping, it is regarded as an input and balances the transmission yield of TCP / UDP packets. To reduce the delay caused by a relatively large specific packet, M is the average packet length or maximum packet length, L is the input packet length, and m 'is 1 or 2 depending on whether M is the average packet length or the maximum packet length. As a value to have, Preferably, when M is an average packet length, the value of 1 is taken, and a maximum packet length is taken.

Is a value corresponding to the processing capacity expressed as the packet loss rate as to how many adjacent packet loss can be compensated for the quality of the voice call in the PLC circuit of the VoIP codec. The voice signal is also calculated for the average of the round trip time (RTT) between the terminal and the gateway and the corresponding session, and applied to the RTT and I values of Equation 3, respectively. The i and j values are time interval values used to obtain the corresponding session number and average. The parameters used in this specification are calculated per class per port, and max _p is applied per class. At this time

The operation takes a plus operation if the average RTT is greater than the individual value, and a minus operation if it does not.

ρ _UT is calculated by Equation 4 as a correlation coefficient between each UDP / TCP packet loss rate (X _U , X _T ) from itself or from a terminal and a previous router node.

From here,

The numerator on the right side is the covariance between UDP packet and TCP packet loss rate,

The right denominator is the product of the standard deviations of the UDP packet and TCP packet loss rates.

During the interval in which the value calculated by Equation 4 is negative and the loss rate of the UDP voice packet is increasing, in the numerator of Equation 3 according to the TCP packet and the UDP voice packet, the former is negative (+) and the latter is reduced ( Perform the operation.

The loss correlation index value indicates a UDP packet loss above the loss limit value for PLC processing in the voice terminal to the upper node directly or through the VoIP server at the terminal, and the VoIP management module 290 receives the RED queue management 244. A request is made to request a drop packet when a silence packet is received.

Hereinafter, referring to FIG. 4, a process of classifying and demultiplexing traffic packets input from the external network for each differential service and performing marking or shaping for congestion control according to each packet type will be described in detail.

FIG. 4 is a diagram schematically illustrating a process in which a packet received in a congestion control device according to another embodiment of the present invention is processed by a classification and demultiplexing module and a marking shaping module.

First, the classifying and demultiplexing module 200 receives a traffic packet 201 input from an external network, and then performs a demultiplexing process by dividing the traffic packet by a differential service (400).

Next, marking and shaping processing is performed on the packet subjected to the demultiplexing process according to the subscriber service and bandwidth allocation according to the user profile of the packet. For this purpose, it is determined whether the VoIP voice packet is a non-mute packet ( 404).

If the VoIP voice packet is a silent packet, it is determined whether or not the set PLC upper limit value corresponding to the loss coping capability of the codec exceeds the threshold x (408). When the network congestion is expected because the PLC upper limit exceeds the threshold x, the AF transmission band allocator 232 performs the safe shaping on the AF packet (420), and if the PLC upper limit does not exceed the threshold x, the AF transmission is performed. The band allocator 232 performs marking on the AF packet.

If the VoIP voice packet is a non-mute packet, it is marked at the premium 238 (412).

Hereinafter, referring to FIG. 5, in a congestion control device according to another embodiment of the present invention, a VoIP management module performs queue control management according to an extended RED algorithm according to the present invention, and also interworks with a PLC helper message. This will be described in detail.

FIG. 5 is a flowchart schematically illustrating an interworking relationship between a PLC helper of a queuing module and a scheduler module and a VoIP management module in a congestion control device according to another embodiment of the present invention.

First, when the VoIP management module 290 receives an operator message or a traffic message from a VoIP support server or a network management system or a packet control related signal message or a traffic control message at a called terminal (500), the message is transmitted to the PLC helper. It is determined whether or not the PLC helper message to be requested (504).

If the message is a PLC helper message, the PLC helper 258 sends an input message 280 from the VoIP management module 290 to the upstream node 258 of the VoIP session every RTCP cycle. In step 512, a PLC helper function of generating and transmitting a control message for the PLC-related information constructed based on the "

If the message received as a result of the determination of the message is not a PLC helper message, request the queue management parameter to the queuing module 240, and the queuing module 240 from the VoIP management module 290. If the requested queue management parameter is a parameter to be performed in its own node, queue management is performed using the extended RED algorithm according to the present invention (508). In this case, if the parameter is updated, the queue control management is performed by updating the RED parameter value reflecting this (520).

Next, the packet processed by the queuing module 240 is transmitted to the scheduler module 250 (520).

The present invention described above is not limited to the stated embodiments and the accompanying drawings, and various substitutions, modifications, and changes are possible in the art without departing from the technical spirit of the present invention. It will be clear to those of ordinary knowledge.

As described above, the present invention provides a voice packet call for congestion in a media gateway having a structure of a module for distinguishing packets according to differential services, a marking and shaping module, a traffic control and queuing module, a scheduler module, and a VoIP management module. By considering the drop rate, queue management is performed for packets in the internal node that meets the PLC processing capacity of the terminal.It controls the congestion in the network and reduces the performance in the PLC during congestion, as well as the traffic congestion in the switching module and network operation. Performance degradation due to parallel link generation can be avoided. In addition, it is possible to automatically provide the end user with the optimal voice call quality suitable for the network situation according to the traffic demand and traffic cancellation as well as user demand. Therefore, it is possible to reduce network bandwidth waste due to the existing RED, and it is possible to operate the network that is resistant to delay and jitter, so that it can perform a reliable packet transmission function as a VoIP gateway or an Internet network node.

Claims (26)

A queuing module for managing queue control for packets received via the VoIP network; A scheduler module including a packet scheduler for transmitting a packet processed by the queuing module to a next node and a PLC helper for creating an RTCP message associated with PLC processing information and sending the packet to an upstream node; And, A congestion control device of a VoIP network including a VoIP management module for receiving a traffic or packet control message input from an external device and analyzing the same by analyzing the queuing module and the PLC helper of the scheduler module. The traffic or packet control related message received by the VoIP management module is one or more messages selected from a message transmitted from an operator, a message transmitted from a VoIP support server or a network management system, or a message transmitted from a called terminal. The PLC helper is based on the input message from the VoIP management module, loss correlation index, round trip time (RTT), loss in PLC circuit of the corresponding codec based on UDP / TCP packet loss rate collected by the VoIP packet receiving end gateway. An apparatus for controlling congestion in a VoIP network, which generates an RTCP message associated with the PLC processing information, including processing capability information and jitter values for predicting congestion in the network, and sends it to an upstream node of an ongoing VoIP session every RTCP cycle. The method of claim 1, A classification and demultiplexing module for performing demultiplexing by dividing the received packets by differential services before the queuing module manages queue control for packets received through a VoIP network; And, And a marking / shaping module for processing marking or shaping of the packets processed by the classifying and demultiplexing module according to the type. The method of claim 2, The classifying and demultiplexing module demultiplexes a TCP source if a traffic packet input from an external network corresponds to a TCP packet, and a non-TP source if the packet does not correspond to a TCP packet. A congestion control device for a VoIP network, characterized in that a demultiplexing process is performed using a TCP source. The method of claim 2, The marking / safety module, for the demultiplexed packet, if the packet is a best packet corresponding to the best service of the IETF differential service, marking or shaping corresponding to a service level agreement (SLA) for each packet in the best service processor. And if the packet corresponds to a band-allocated packet of an Expedited Forwarding (EF) band guarantee service, a marking or shaping process corresponding to a service level agreement (SLA) for each packet is performed at a premium device, or the packet In the case of a packet corresponding to the AF (Assured Forwarding) service of this differential service, the congestion of the VoIP network, wherein the AF transmission band allocator performs marking or shaping processing corresponding to a service level agreement (SLA) for each packet. controller. The method of claim 3, wherein The marking / shaping module performs demultiplexing processing on the demultiplexed packet through the BTC source device of the classification and demultiplexing module, and when the packet is a non-mute packet, marks the packet at a premium device, and the packet is a silent packet. In this case, when the set PLC upper limit value corresponding to the loss coping force of the codec exceeds the threshold value, the AF transmission band allocator performs the safe processing, and when the packet is a silent packet, the PLC upper limit value exceeds the threshold value. Otherwise, the apparatus for controlling congestion of the VoIP network, wherein the AF transmission band allocator performs marking processing on the AF packet. The method of claim 5, The demultiplexed packet through the BTP source device is a congestion control device of a VoIP network, characterized in that the VoIP voice packet. The method of claim 5, The marking is congestion control device of the VoIP network, characterized in that using the IETF Differentiated Service Code Point (DSCP) code point or Type of Service (ToS) bits of the packet header. The method of claim 4, wherein The queuing module may include a RED queue for storing queued AF packets input from a marking / shaping module and performing queue control management according to a RED algorithm; And, And a FIFO queue configured to store the best packet inputted from the best service processor of the marking / safety module and perform a corresponding queue management algorithm. The method of claim 8, The RED queue of the queuing module includes a first queue management algorithm that performs packet drop based on an average queue length estimation according to a parameter transmitted from the VoIP management module, and a terminal-side VoIP packet in consideration of the limitation of packet loss compensation in the VoIP service. And a traffic / label controller configured to manage queue control by selectively using a second queue management algorithm that performs packet drop by using the processing information. The method of claim 9, The first queue management algorithm performed in the traffic / label controller is a congestion control device for a VoIP network, characterized in that the packet drop is performed for the RED queue by applying the packet drop probability calculated using Equation 2 below. [Equation 2]

From here, S _n is the average queue length, k _l is the lower average queue length, k _h is the upper average queue length, max _p is the maximum drop probability value between k _l and k _h . The method of claim 10, The average queue length S _n is a congestion control device of a VoIP network, characterized in that calculated using the following equation (1). [Equation 1]

From here, S _n is the average queue length, w is the weight applied to the current queue length, q _n is the current queue length, m is the system input load throughput value. The method of claim 9, The second queue management algorithm performed by the traffic / label controller is a congestion control device of the VoIP network, characterized in that the packet drop is performed for the RED queue by applying the packet drop probability calculated using Equation (3). [Equation 3]

From here, Pd is the packet drop probability calculated according to the first queue management algorithm, N is the number of input packets since the last drop from the node performing RED queue management, M is the average packet length or maximum packet length, L is the input packet length, m 'is 1 if M is the average packet length, 2 if the maximum packet length,

Is a value corresponding to the processing capacity expressed as the packet loss rate as to how many adjacent packet loss can be compensated for the required quality of voice call in PLC circuit of VoIP codec, RTT _j and I _i are the average of the round trip time (RTT) between the terminal and the gateway for the voice signal and the corresponding session. I and j are the time interval values used to obtain the session number and the average,

Operation is a plus operation if the average RTT is greater than an individual value, otherwise a minus operation, ρ _UT is the correlation between each UDP / TCP packet loss rate (X _U , X _T ) from the gateway itself or from the terminal and the previous router node. The method of claim 12, Ρ _UT is a congestion control device of the VoIP network, characterized in that the value calculated by the following equation (4). [Equation 4]

From here, The numerator on the right side is the covariance between UDP packet and TCP packet loss rate, The right denominator is the product of the standard deviations of the UDP packet and TCP packet loss rates. A router comprising the device of any one of claims 1 to 13. A router switch comprising the device of any one of claims 1 to 13. A VoIP gateway comprising the device of any one of claims 1 to 13. A queuing step of queuing a packet received via the VoIP network; A packet scheduling step of transmitting the queued packet to a next node; And, The VoIP management module receives and analyzes a traffic or packet control related message, which is one or more messages selected from a message transmitted from an operator, a message transmitted from a VoIP support server or a network management system, or a message transmitted from a called terminal. A congestion control method of a VoIP network comprising a VoIP management step of managing a step and a packet scheduling step, In the VoIP management step, if the message is a PLC helper drive request message as a result of analyzing the traffic or packet control related message, loss correlation based on UDP / TCP packet loss rate collected by the VoIP packet receiving end gateway from the PLC helper Creates an RTCP message associated with PLC processing information, including exponents, round trip time (RTT), loss processing capability information in the PLC's circuitry of the codec, and jitter values for prediction of congestion in the network, and sends it to the upstream node of an ongoing VoIP session. Send out every RTCP cycle, In the queuing step, a packet drop is performed using a first queue management algorithm that performs packet drop based on an average queue length estimation for an externally transmitted VoIP packet or terminal-side VoIP packet processing information in consideration of a limitation of packet loss compensation in a VoIP service. The congestion control method of the VoIP network which manages queue control selectively using the 2nd queue management algorithm which performs the following. The method of claim 17, A classification and demultiplexing step of performing demultiplexing by dividing the received packets by differential services before the queuing step; And, And a marking / shaping step of processing marking or shaping according to the type of each packet processed in the classifying and demultiplexing step. The method of claim 18, In the classifying and demultiplexing step, if the traffic packet inputted from the external network corresponds to a TCP packet, the demultiplexing process is performed by a TCP source. A congestion control method for a VoIP network, characterized in that a demultiplexing process is performed using a TCP source. The method of claim 18, In the marking / shaping step, if the packet is the best packet corresponding to the best service of the IETF differential service, the marking or shaping corresponding to the service level agreement (SLA) for each packet is performed by the best service processor. And if the packet corresponds to a band-allocated packet of an Expedited Forwarding (EF) band guarantee service, a marking or shaping process corresponding to a service level agreement (SLA) for each packet is performed at a premium device, or the packet In the case of a packet corresponding to the AF (Assured Forwarding) service of this differential service, the congestion of the VoIP network, wherein the AF transmission band allocator performs marking or shaping processing corresponding to a service level agreement (SLA) for each packet. Control method. The method of claim 19, In the marking / shaping step, the demultiplexed packet is demultiplexed through the BTP source device of the classification and demultiplexing step. In the case where the set PLC upper limit value corresponding to the loss coping force of the codec exceeds the threshold value, the AF transmission band allocator performs the safe processing, and when the packet is a silent packet, the PLC upper limit value exceeds the threshold value. Otherwise, the AF transmission band allocator performs a marking process on the AF packet. The method of claim 21, Congestion control method of the VoIP network, characterized in that the demultiplexed packet through the VoIP source device is a VoIP voice packet. The method of claim 17, The first queue management algorithm is a congestion control method of the VoIP network, characterized in that the packet drop is performed for the RED queue by applying the packet drop probability calculated using Equation 2. [Equation 2]

From here, S _n is the average queue length, k _l is the lower average queue length, k _h is the upper average queue length, max _p is the maximum drop probability value between k _l and k _h . The method of claim 23, wherein The average queue length S _n is calculated using the following equation (1) congestion control method of a VoIP network. [Equation 1]

From here, S _n is the average queue length, w is the weight applied to the current queue length, m is the system input load throughput value; The method of claim 17, The second queue management algorithm is a congestion control method of the VoIP network, characterized in that the packet drop is performed for the RED queue by applying the packet drop probability calculated using Equation 3 below. [Equation 3]

From here, Pd is the packet drop probability calculated according to the first queue management algorithm, N is the number of input packets since the last drop from the node performing RED queue management, M is the average packet length or maximum packet length, L is the input packet length, m 'is 1 if M is the average packet length, 2 if the maximum packet length,

Is a value corresponding to the processing capacity expressed as the packet loss rate as to how many adjacent packet loss can be compensated for the required quality of voice call in PLC circuit of VoIP codec, RTT _j and I _i are the average of the round trip time (RTT) between the terminal and the gateway for the voice signal and the corresponding session. I and j are the time interval values used to obtain the session number and the average,

Operation is a plus operation if the average RTT is greater than an individual value, otherwise a minus operation, ρ _UT is the correlation between each UDP / TCP packet loss rate (X _U , X _T ) from the gateway itself or from the terminal and the previous router node. The method of claim 25, Ρ _UT is a congestion control method of a VoIP network, characterized in that the value calculated by the following equation (4). [Equation 4]

From here, The numerator on the right side is the covariance between UDP packet and TCP packet loss rate, The right denominator is the product of the standard deviations of the UDP packet and TCP packet loss rates.

Images