KR100674329B1 - Method for Congestion Control of the Router in TCP/IP - Google Patents

Abstract

The present invention relates to a congestion control method of a router in a transmission control protocol / internet protocol network. When a network congestion occurs, a packet received from a router uses a TCP protocol to allocate a fair network resource to each information source regardless of a transport layer protocol. Packets of adaptive information sources are divided into packets of non-adaptive information sources using UDP protocol, and the packet discard method and service rate are applied differently according to the packets. Accordingly, the TCP information source can be protected against unintentional or intentional attack by the UDP information source, provide a reliable network environment to the user, and have the advantage of identifying the correct network state.

Description

Method for Congestion Control of the Router in TCP / IP in Transmission Control Protocol / Internet Protocol Networks

1 is a block diagram of a TCP packet header in a TCP / IP network to which the present invention is applied.

2 is a block diagram of a UDP packet header in a TCP / IP network to which the present invention is applied.

3 is a diagram illustrating a packet discard method using a conventional RED algorithm.

4 is a flowchart illustrating a packet discard process according to FIG. 3.

5 is a view showing a packet discard method according to the present invention.

6 is a flowchart illustrating a packet service process according to FIG. 5.

<Description of the symbols for the main parts of the drawings>

10: TCP packet header 11, 31: source port

12,32: destination port 13: sequence number

14: Acknowledgment Number 15: Windows

16,34: check sum 17: agent pointer

18,35: data 30: UDP packet header

33: UDP message length 50,102: buffer

101: classification unit 102a, 102b: queue

104: scheduler

The present invention is fair regardless of the transport layer protocol of the user data when traffic congestion occurs in a data communication network based on a transmission control protocol / Internet protocol (hereinafter referred to as TCP / IP). In order to allocate resources, more specifically, the transmission control protocol / Internet protocol that separates data into adaptive information sources (Non-Adaptive and Non-Adaptive information sources) and stores them in each buffer by applying different packet discard algorithms. The present invention relates to a congestion control method of a router in a network.

The Internet started as a non-profit information and communication network connecting information media around the world, but recently developed into a core infrastructure of the future information society following the explosive increase in Internet traffic caused by the rapid growth of PC communication and the World Wide Web. Doing.

In addition, asynchronous transfer mode (ATM) communication network, which is attracting attention as the information and communication network of the future, has developed very little application service using native ATM API in the world, and the commercial use of the internet is emerging as a major issue. .

Therefore, for the time being, all application services are expected to use TCP / IP based Internet.

TCP / IP layer is composed of Application Layer, Transport Layer, Internet Layer, and Physical Layer. Currently, widely used transport layer protocols are considered to be based on network conditions, such as TCP. There are protocols for controlling data and protocols for transmitting data unconditionally without considering network conditions, such as User Datagram Protocol (hereinafter referred to as UDP).

1 is a configuration diagram of a TCP packet header 10 in a TCP / IP network to which the present invention is applied, and shows a structure of a TCP packet used by a TCP information source for transmitting data in segments.

1 is a source port 11 indicating an address of a source (sender) for transmitting a TCP packet;

A destination port 12 indicating an address of a destination (destination) to receive the TCP packet;

A sequence number 13 indicating a data transmission start position in the entire message (data) stream;

An acknowledgment number 14 indicating a data position to be transmitted next in the entire message (data) stream;

Sliding-Window that uses a Retransmission Timeout (RTO) and Duplicate Ack to send a certain number of packets consecutively between the sending and receiving sides, and retransmits the packets if they do not reach their destination. A window 15 for flow control;

A checksum 160 indicative of the number of bits in the transmission unit to determine whether the same number of bits have arrived at the destination; and

An urgent pointer 17 indicating the size of urgent data in the segment; And

It consists of Data 18 representing the data to be transferred.

The TCP information source determines the amount of data that can be transmitted at one time by using the size of the window 15 when transmitting data, and implicitly detects the traffic congestion of the network by using the retransmission timeout and the replication response signal, according to the network state. Adjust data transfer.

That is, if the acknowledgment message does not arrive during the retransmission timeout calculated by the round trip time (RTT), the size of the window 15 is reduced by considering that congestion has occurred in the network.

Therefore, the TCP window size decreases exponentially with packet loss due to network congestion.

2 is a diagram illustrating a structure of a UDP packet header in a TCP / IP network to which the present invention is applied, and shows a structure of a UDP packet used by a UDP information source.

2 shows a source port 31 indicating an address of a source (sender) for transmitting a UDP packet;                         

A destination port 32 indicating an address of a destination (destination) to receive the UDP packet;

A message length 33 indicating the total number of data bits of the UDP packet;

A check sum 34 for confirming whether the same number of bits have arrived at the destination using the number of bits in the transmission unit; And

It consists of data 35 representing the data to be transmitted.

The structure of the UDP packet header is simpler than that of the TCP packet header shown in FIG. 1, and does not have information for adjusting the transmission rate depending on reliable transmission of data or network conditions.

Until now, the Internet service provides only the best data-oriented service that does not require strict service quality between end-to-end, but since the Internet is developing as an infrastructure for providing multimedia service, it can satisfy various service quality requirements of application services. The development of the Internet is required.

Since these multimedia services require the transmission of data in real time, the UDP protocol is preferred to the TCP protocol as the transport layer protocol. Recently, as users of network games, audio, and video are increasing, traffic using UDP as a transport layer protocol is rapidly increasing.

The information source using the UDP protocol continuously generates traffic and transmits it to the network without considering the state of the network, thereby increasing the network congestion and reducing the performance of the TCP connection.                         

In addition, non-adaptive information sources using the UDP protocol may intentionally or unknowingly damage other sources using the TCP protocol due to their aggressive nature.

Therefore, there is a need for a router with a large queue in preparation for temporary traffic congestion in a high speed network with a large bandwidth connection.

However, having a large queue in the current network environment where only TCP flow control exists is undesirable because it increases the average delay and reduces the throughput.

Transport layer protocol estimates network congestion from service time, yield change, end-to-end (eg end to end, user to user, etc.) delay, packet loss, etc. in network environment without feed-back control. can do.

However, because network congestion is affected by several factors, such as the number of congested routers, traffic patterns in connections, changes in routing, and delays in transmission due to queuing delays, the transport layer protocol for each connection Difficult to determine judgment or cause.

The detection of network congestion and the control of the extent and duration of congestion can be most effectively performed in the router. In the router, the connection traffic processing method all the networks share the delay during the congestion, It is handled in a first-in, first-out (FIFO) manner.

Currently, the packet discard algorithm RED (Random Early Detection) is widely used for congestion control in the Internet gateway or router.

3 is a diagram illustrating a packet discard method using a conventional RED algorithm, and illustrates a packet discard method in a RED buffer 50 having a minimum threshold and a maximum threshold.

FIG. 3 shows that the packet discard probability varies from "0" to "max_p" between the minimum threshold value and the maximum threshold value maxthresh_. The equation for determining the packet discard probability Probl_ is described below. Equation 1

[Equation 1]

if (bytes_), pl = max_p (ave_-thresh _) / (maxthresh_-thresh_)

            pl = pl (pksize / Max_pksize_)

            probl_ = pl / (1-count pl)

In Equation 1, count is the number of packets stored in the buffer 50 after discarding packets belonging to an arbitrary flow, and ave_ represents an average queue size, which is an exponential weighted moving average (EWMA). It calculates using and the calculation formula is the same as Formula 2 described below.

[Equation 2]

ave_ = (1-q_weight_) * ave_ + q_weight_ * curq_

FIG. 4 is a flowchart illustrating a packet discard process according to FIG. 3, and illustrates a process of determining whether to discard a packet received at a router or a gateway according to an average queue size.                         

First, a packet is received from an information source (for example, an adaptive information source or a non-adaptive information source) (S2).

The average queue length Qa of the current queue is calculated (S4), and it is determined whether the average queue length Qa is larger than the minimum threshold threshold_ (S6).

As a result of the determination, if it is smaller than the minimum threshold threshold_, the "count" variable is increased (S22) and stored in the RED buffer 50 (S18).

As a result of the determination, if it is greater than the minimum threshold thresh_, it is determined whether it is smaller than the maximum threshold maxthresh_ (S8).

As a result of the determination, if the maximum threshold value (maxthresh_) is larger than the "count" variable value is "0" (S20), the packet is discarded (S22).

As a result of the determination, if it is smaller than the maximum threshold maxthresh_, the packet discard probability value Pa is calculated by Equation 1 (S10), and a random number R is generated using the random number generator (S12).

It is determined whether the packet discard probability value Pa is smaller than the random number R (S14). If the packet probability probability Pa is small (S14), the value of the "count" variable is increased (S16), and the packet is buffered. Store in (S18).

As a result of the determination, if the packet discard probability value Pa is greater than the random number R, the variable "count" is made "0" (S20), and the packet is discarded (S22).

As shown in FIG. 4, the RED packet algorithm monitors a queue and randomly selects a connection and discards a packet belonging to the connection when the average queue size Qa exceeds a threshold. In this case, the probability that a connection is selected is proportional to the traffic amount of the input connection.

Therefore, the RED packet discard algorithm can be performed gradually without being affected by the congestion control method in other routers of the network regardless of the transport layer protocol, and control the average queue size to prevent congestion and synchronize the information source of TCP. It solves Global Synchronization and maintains an upper limit on the average queue size even when there is no flow control of the transport layer protocol.

Routers using the RED packet discard algorithm can easily control the average queue size when the transport layer does not use traffic flow control or when rate flow control is used.

Basically, however, the RED algorithm is designed for networks where marked or discarded packets can inform the transport layer protocol of network congestion, which is useful for TCP-only application services, but for nodes with mixed UDP and TCP traffic (e.g., Routers, gateways, etc.) have their limitations.

That is, when congestion occurs, the RED algorithm discards the packet based on the same criteria for both the TCP packet and the UDP packet. As a result of the packets discarded by the RED algorithm, the TCP information source reduces the amount of packet transmission, but the UDP information source transmits the packet as much as possible if there is data to be transmitted regardless of the discarded packet.

Therefore, there is a disadvantage in that the amount of data transmitted by the UDP source for a single time is much larger than the amount of data transmitted by the TCP source.                         

For this reason, with the increase of the application services using the UDP transport protocol, the weight of the traffic using the UDP protocol in the network is increasing, and in the worst case, the packet transmission of the TCP information source is gradually decreasing, and the UDP does not consider the network state. Packet transmission of information sources is increasing, which has the disadvantage of unfairly using network resources.

Accordingly, in view of the above-mentioned problems, the present invention provides a packet received from a router to allocate fair network resources to each information source regardless of a transport layer protocol. The purpose of this section is to provide a congestion control method of a router in a transmission control protocol / Internet protocol network that classifies packets into non-adaptive information sources using UDP protocol and applies packet discard methods and service rates differently according to the packets.

In the transmission control protocol / internet protocol network, the congestion control method of the router in the transmission control protocol / Internet protocol network, in the method of controlling the network congestion of the router in the transmission control protocol / Internet protocol network, when a packet is received, enters the protocol field of the packet. Classifying the packet into an adaptive packet or a non-adaptive packet using a stored value; A second step of storing an adaptive packet in an adaptive storage and a non-adaptive packet in a non-adaptive storage by a packet discard algorithm according to the type of the packet; A third step of calculating an adaptive traffic service rate and a service rate of non-adaptive traffic according to the traffic type; And a fourth process of transmitting the packet according to a service rate obtained by adding up the adaptive traffic service rate calculated by the third process and the service rate of non-adaptive traffic.

delete

delete

delete

delete

delete

The above objects, features and advantages will become more apparent from the following detailed description taken in conjunction with the accompanying drawings. Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

FIG. 5 is a diagram illustrating a packet discard method according to the present invention. When a packet is received by a router, the received packet is divided into adaptive queues 102a and non-adaptive queues 102b according to a transport layer protocol, and stored in a queue state. Shows how to adjust the packet service rate.

5 is a classification unit 101 for classifying an adaptive packet and a non-adaptive packet with reference to a protocol field value of a received packet;

A buffer (100) having a queue structure consisting of an Adaptive queue (102a) for storing the Adaptive packet and a Non-Adaptive queue (102b) for storing Non-Adaptive packets; And

It is composed of a scheduler 104 for adjusting the packet service rate according to the state of the queue (102a ~ 102b).

The protocol field in the IP header of the packet has a unique field value for each transport layer protocol. The field values are shown in Table 1 below.

TABLE 1

protocol Internet control message protocol (ICMP) Internet Group Management Protocol (IGMP) TCP UDP Field value One 2 6 17

The classifying unit 101 classifies an adaptive packet and a non-adaptive packet with reference to the transport layer protocol information (field value) shown in Table 1, and classifies it as an adaptive packet when the field value is "6" (TCP), Since the Internet Control Message Protocol and the Internet Group Management Protocol do not consider the state of the network during traffic transmission, such as UDP, packets with field values other than "6" are classified as non-adaptive packets.

The sizes of the queues 102a to 102b are set in consideration of flow characteristics such as packet delay and packet loss. Since non-adaptive packets have a small burst property, the size of the non-adaptive queue 102b is increased. Doing so degrades performance.

In addition, the management method of each of the queues 102a to 102b is different from that of the traffic transmission method of the adaptive information source and the non-Adaptive information source. Therefore, the management method is also applied differently.

Accordingly, the adaptive queue 102a manages by applying the RED packet discard algorithm of FIG. 3 having a threshold in consideration of the burst degree of traffic, and the non-Adaptive queue 102b receives the received packet when the storage space of the queue is full. It is managed using a DropTail packet dropping algorithm that discards.

The scheduler 104 actively adjusts the service rates of the queues 102a to 102b by using weights considering the states of the queues 102a to 102b and the characteristics of the packets stored in the queues.

The scheduler 104 measures the input traffic and calculates an average value (or median value) of the input traffic amount for each of the queues 102a to 102b, and compares the inverse of the length of the queues 102a to 102b and the current service rate. The service rate is adjusted by calculating a new proportional service rate for each queue.

Algorithm 1 described below is an algorithm for calculating the service rate, which is periodically updated.

Algorithm 1

Constants:

Link _B : Llink Bandwidth (LinkK _B = SR-Min _N + SR-Max _A = SR-Max _N + SR-Min _A )

SR-Min _N : Minimum service rate for the Non-Adaptive buffer

SR-Max _N : Maximum sevice rate for the Non-Adaptive buffer

SR-Min _A : Minimum service rate for the Adaptive buffer

SR-Max _A : Maximum sevice rate for the Adaptive buffer

QT-Min _N : Minimum threshold of the Non-Adaptive queue

QT-Max _N : Maximum threshold of the Non-Adaptive queue

QT-Min _A : Minimum threshold of the Adaptive queue

QT-Max _A : Maximum threshold of the Adaptive queue

W ₁ : Weight for the calculation of the EWMA queue length

W ₂ : Weight for the calculation of the EWMA service rate

Variables:

SR _N : service rate of the Non-Adaptive traffic for a period. (SR-Min _N ≤ SR _N ≤ SR-Max _N )

SR _A : service rate of the Adaptive traffic for a period. (SR-Min _A ≤ SR _A ≤ SR-Max _A )

AQL _N : Average Queue Length of the Non-Adaptive buffer for a period.

AQL _A : Average Queue Length of the Adaptive buffer for a period.

Q: current queue size

AQL: average queue size

// Initializing                     

IncRatePerByte _N = (SR-Max _N -SR-Min _N ) / (QT-Max _N -QT-Min _N )

IncRatePerByte _A = (SR-Max _A -SR-Min _A ) / (QT-Max _A -QT-Min _A )

Enque Part:

// Classify the packet into Adaptive packet and Non-Adaptive packet.

if (P ∈ Adaptive flows) {

 Store P into the Adaptive buffer

 Implement RED algorithm

}

else {

 Store P into the Non-Adaptive buffer

 Implement Drop-Tail algorithm

}

// Calculate the average size of each queue

Calculate the average queue size during the period.

if (start the new period)

AQL = 0

if (Q ∥packet entered)

AQL = (1-W ₁ ) * AQL + W ₁ * Q

// Calculate the service rate for each queue                     

Calculate the service rate for each queue.

// Process of calculating the service rate of non-adaptive queue.

// Calculate the service rate for Non-Adaptive packets.

if (AQL _N > QT-Min _N ) {

temp = ((AQL _N -QT-Min _N ) * IncRatePerByte _N ) + SR-Min _N

SR _N = (1-W ₂ ) * SR _N + W ₂ * temp

}

else

SR _N = SR _N -Min _N

// Service rate calculation process of adaptive queue

// Caculate the service rate for the adaptive packets.

if (AQL _A > QT-MIN _A ) {

temp = ((AQL _A -QT-Min _A ) * incRateByte _A ) + SR-MIN _A

SR _A = (1-W ₂ ) * SR _A + W ₂ * temp

}

else

SR _A = SR-Min _A

The process of calculating the new service rate by Algorithm 1 is the same for both adaptive traffic and non-adaptive traffic. First, the amount of traffic for each queue is calculated. Calculate each average queue size for adaptive traffic.

If the calculated average queue size is larger than the minimum threshold of the queue, each new service rate for adaptive traffic and non-adaptive traffic is calculated in proportion to the inverse of the queue size and the current service rate of the corresponding traffic.

If the calculated average queue size is smaller than the minimum threshold of the queue, the minimum service rate of the current traffic is calculated as the new service rate for adaptive traffic and non-adaptive traffic.

As such, the new service rate calculation for adaptive traffic and non-adaptive traffic is estimated by considering the buffer status in addition to the minimum and maximum traffic amounts determined by estimating the traffic volume of each application service.

FIG. 6 is a flowchart illustrating a packet service process according to FIG. 5. The packet received by the router is classified into an adaptive packet and a non-adaptive packet according to its characteristics, and according to algorithms and service rates of the corresponding queues 102a to 102b. Indicates the process of providing a service.

First, a packet is received from a user (S100).

The protocol field value of the received packet is referred to (S102), and it is determined whether the field value is equal to "6" (TCP packet) (S104).

As a result of the determination, if the field value is "6", the RED packet discard algorithm is applied (S106), and whether the packet is stored or not is determined by the RED packet discard algorithm and stored in the adaptive queue S02a if appropriate. If not, the packet is discarded (S108).

A new service rate is calculated for each of the queues 102a to 102b (S110), and the packet service is provided at a rate of adding up the service rate of the calculated adaptive queue and the non-Adaptive queue (S112).

As a result of the determination, if the field value is not "6", the Drop-Tail packet discard algorithm is applied (S114), and if the storage of the packet is determined by the RED packet discard algorithm, and if appropriate, Non-Adaptive queue 102a. If not stored in the packet is discarded (S116).

A new service rate is calculated for each of the queues 102a to 102b (S110), and the packet service is provided at a rate of adding up the service rate of the calculated adaptive queue and the non-Adaptive queue (S112).

As described above, according to the present invention, there is an advantage of allocating fair resources to the information source using the TCP protocol and the information source using the UDP protocol in the TCP / IP network by managing the TCP and UDP packets separately in the router during network congestion. have.

In addition, the TCP information source can be protected against unintentional or intentional attack by the UDP information source, provide a reliable network environment to the user, and have the advantage of identifying the exact network state.                     

In addition, preferred embodiments of the present invention are disclosed for the purpose of illustration, those skilled in the art will be able to various modifications, changes, additions, etc. within the spirit and scope of the present invention, such modifications and modifications belong to the following claims You will have to look.

Claims (7)

In the control method of network congestion of a router in a transmission control protocol / Internet protocol network, When the packet is received, classifying the packet into an Adaptive packet or a Non-Adaptive packet using a value stored in a protocol field of the packet; A second step of storing an adaptive packet in an adaptive storage and a non-adaptive packet in a non-adaptive storage by a packet discard algorithm according to the type of the packet; A third step of calculating an adaptive traffic service rate and a service rate of non-adaptive traffic according to the traffic type; And And a fourth step of transmitting the packet according to a service rate obtained by adding up the adaptive traffic service rate calculated by the third process and the service rate of non-adaptive traffic. Control method of congestion. The method of claim 1, In the first step, classifying a packet according to a traffic type may include: Referencing a protocol field value of a received packet; Classifying the packet as an adaptive packet if a protocol field value of the packet is equal to a field value indicating a TCP protocol through protocol field information of an embedded packet; And And classifying the packet into a non-adaptive packet if the protocol field value of the packet is not the same as the field value indicating the TCP protocol. The method of claim 1, The packet storing method by the packet discard algorithm of the second step, If the packet is an Adaptive packet, storing the packet in an adaptive storage unit using a RED packet discard algorithm; And And storing the packet in a non-adaptive storage unit using a drop-tail packet discard algorithm if the packet is a non-adaptive packet. The method of claim 3, Adaptive packet storage using the RED packet discard algorithm, Calculating an average length of a queue of the current adaptive storage; If the average length is less than the minimum threshold of the queue, storing the packet in Adaptive storage; When the average queue length is larger than the minimum threshold of the queue and smaller than the maximum threshold, the packet discard probability value is compared with the random number generated by the random number generator. If the packet discard probability value is smaller than the random number, the packet is adaptively stored. Storing in the unit and discarding the packet if a packet discard probability value is greater than the random number; And And discarding the packet if the average queue length is greater than the maximum threshold of the queue. 2. The method of controlling congestion of a router in a transmission control protocol / internet protocol network. The method of claim 3, The storing of the non-adaptive packet using the drop-tail packet discard algorithm, Checking whether there is free space for storing the packet in a queue of the non-adaptive storage unit; Storing the packet in a non-adaptive storage unit when there is free space for storing the packet in the queue; And discarding the packet when there is no free space to store the packet in the queue. The method of claim 1, Computing the service rate of the adaptive storage unit in the third process, Calculating a traffic amount of the adaptive storage unit; Calculating an average queue size of the adaptive storage unit according to the traffic amount; If the calculated average queue size is larger than the minimum threshold of the Adaptive queue, calculating a service rate of Adaptive traffic in proportion to the inverse of the size of the Adaptive storage and the current service rate; And When the calculated average queue size is smaller than the minimum threshold of the adaptive queue, calculating the minimum service rate of the adaptive storage as the service rate of the adaptive traffic; and controlling the congestion of the router in a transmission control protocol / internet protocol network . The method of claim 1, In the third process, the process of calculating the service rate of the non-adaptive storage unit may include: Calculating a traffic amount of the non-adaptive storage unit; Calculating an average queue size of the non-adaptive storage unit according to the traffic amount; Calculating a service rate of the non-adaptive traffic in proportion to the inverse of the size of the non-adaptive storage and the current service rate when the calculated average queue size is larger than the minimum threshold of the non-adaptive queue; And If the calculated average queue size is less than the minimum threshold of the Non-Adaptive queue, calculating the minimum service rate of the non-Adaptive storage as the service rate of the non-Adaptive traffic, characterized in that the transmission control protocol / Internet Protocol network Control of congestion in the router

Images