KR100494562B1 - Class based TCP congestion control method - Google Patents

Abstract

The present invention relates to data transmission, and more particularly, to a transmission control protocol (TCP) congestion control method. In the present invention, using a class-based TCP congestion control method that is differentiated in a transmission node rather than a network node device, in the case of a real-time service such as an identity bit rate service, the narrowing conditions of the band are more stringent, and a relatively non-real time service or a variable bit rate is used. Bandwidth concessions shall be made for the services of. To this end, the extent of transmission window reduction varies depending on the type of service during network congestion. By using the method of the present invention, it is possible to provide a differentiated service according to the type of service without modifying or replacing network equipment by being compatible with existing equipment.

Description

Class based TCP congestion control method

The present invention relates to data transmission, and more particularly, to a transmission control protocol (TCP) congestion control method.

Recently, as the necessity and demand of the storage area network (SAN) increases, it replaces the existing Fiber Channel-SAN (FC-SAN), which is incompatible with network equipment. Internet Protocol-SAN (IP-SAN) technology has emerged. Using iSCSI, SANs can be configured to transfer information between servers and storage over IP networks.

SAN is a network technology that makes it easy to share and access data and resources by providing access to multiple Fiber Channel-connected storage using a hub or switch.

iSCSI technology is a standard created by the Internet Engineering Task Force (IETF) that allows users to access the storage they want by encapsulating the SCSI interface, which is used to exchange data with storage on the SAN, and routing it to the IP network by encapsulating it in TCP. . This specification specifies that only TCP is used. The data between the server and the storage has various characteristics from web data delivered in real time to video on demand (VOD) data delivered in real time, and user requirement information.

However, TCP is effective for non-real-time services, but there are various problems in applying directly to real-time services. Conventional congestion control method of TCP improves congestion by reducing the TCP transmission window size to 1/2 when loss of the transmitted segment occurs. This method is widely used as a suitable method for non-real-time services. However, even in the case of a real-time service having a constant bitrate characteristic, if the TCP transmission window size is reduced to 1/2 when a segment is lost, the bandwidth reserved for providing the constant bitrate service cannot be guaranteed.

Differentiated Service (DiffServ) is a protocol for classifying and controlling network traffic by class so that it has priority over other types of general traffic for relatively special types of traffic such that voice should not be interrupted. to be. Therefore, by using the differential service (DiffServ), it is possible to differentiate by class using a method such as Class Based Queuing in the intermediate node device. Here, the class means classification according to the characteristics of the service, and the class requiring the same characteristics of the service in DiffServ is defined as one class. However, in order to make such a difference, there is a problem such as modifying or replacing existing equipment.

 The technical problem to be achieved by the present invention is to provide a differentiated service according to the characteristics of the Internet data transmitted and received without modifying or replacing the existing network intermediate node equipment using the TCP congestion control algorithm of the transmitter and the user's needs. To this end, when congestion of the network is detected due to the loss of the transmitted TCP packet, a TCP congestion control method is provided to reduce the transmission window size to escape from the congestion state.

In order to achieve the above object, a method for controlling congestion of TCP traffic according to the present invention includes: determining whether TCP data to be transmitted to the Internet network is real-time service data or non-real-time service data; And adjusting the reduction degree of the transmission window size of the TCP data differently according to the type of the service.

In order to achieve the above object, a method of controlling congestion of TCP traffic according to the present invention includes a performance of a real-time service class network in which TCP data of a real-time service is transmitted and a non-real-time service class network in which TCP data of a non-real-time service is transmitted. Setting a relative performance indicating a ratio with the performance to a value desired by the user; Recalculating a transmission window reduction factor of the real-time service class network or a transmission window reduction factor of the non-real-time service class network according to the set relative performance value; And adjusting the TCP data transmission window size using the calculated transmission window reduction factor according to the type of TCP data to be transmitted.

In order to achieve the above object, the present invention provides a computer-readable recording medium recording a program for executing the method on a computer.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

The conventional TCP congestion control method solves the congestion situation by reducing the size of the transmission window when congestion is detected by packet loss. At this time, the reduction ratio of the transmission window size is reduced to a certain ratio (1/2) regardless of the service type. However, the present invention improves the transmission window size reduction method by discriminating according to the type of service. In the present invention, as a type of service, a class requiring a real time service and a service class without a real time service will be described as an example.

When the reduction factor of the transmission window size in the TCP congestion control is Δ, the network performance B [P] of the round trip time (RTT) and the loss rate P may be expressed by Equation 1 below.

Here, the round trip time (RTT) refers to the time taken for one packet to make a round trip between both ends, and b is This means that after receiving several packets, an acknowledgment signal (ACK) is sent. TCP sends an acknowledgment signal (ACK) to the received packet, which may send an ACK every time it receives (b = 1) or ACK after receiving n packets (b = n). And p is the rate of packet loss, which is usually lost when the network is congested and unable to send all packets, which is the rate of loss. A large value of B [p] calculated in this way means that the network performance is good. The use of the meaning will be described below.

Equation 1 may be used to predict network performance for a general reduction factor Δ. Referring to Equation 1, it can be seen that the network performance can be adjusted by adjusting the reduction factor Δ. However, adjusting the reduction factor Δ as a way to solve the congestion situation in a situation where limited resources are required cannot achieve the improvement of the overall network performance. This is because variables such as loss rate P are not independent of reduction factor Δ. In other words, the maximum performance can be obtained only by adjusting Δ when the loss rate P is not affected by the reduction factor Δ. Since the loss rate P may be affected by the reduction factor Δ, the adjustment of relative Δ relative to other traffic rather than the absolute adjustment of Δ It is necessary to achieve predictable performance improvement.

Therefore, the adjustment of the reduction factor Δ is useful for adjusting relative performance. Referring to Equation 2, it can be seen the relative performance difference between the performance ratio R between the classes (class) that uses a class (class) and Δ ₂ Δ ₁ for use in the same network. That is, Equation 2 calculates a performance ratio between two classes when each class in Equation 1 has the same RTT, b, p.

While the conventional method uses only the reduction factor Δ = 2, the present invention derives a difference in transmission performance by using different reduction factors of Δ ₁ and Δ ₂ for two kinds of service classes. Using Equation 3, it can be a reduction factor compared to the class that is used to obtain the Δ ₁ Δ ₂ has a relatively reduced factor of R times the performance. That is, if both sides of Equation 2 are squared and summed about Δ ₂ , Equation 3 can be obtained.

Even if the reduction factors Δ ₁ and Δ ₂ for each class calculated by Equations 1 to 3 are applied, it is necessary to maintain characteristics similar to a network using only Δ = 2 because it uses limited resources. If the ratio of relative performance between classes is relative performance, using Equation 4, the composition ratio of traffic for each class can be obtained to satisfy the relative performance R _ref when Δ = 2 is used.

Here, X means a class of real-time traffic characteristics, Y means a class of non-real-time characteristics. In more detail, the relative performance means relative performance of two classes, and one way of using the equation of the equation 1 using the same network path, that is, the same RTT, b, p The relative performance ratio of two services of different classes applied. Thus, R ₁ , R ₂ is capable of each class represented by the first means and the relative performance ratio of class 2 and one R ₁ ratio based on the ratio of the R ₂ or R ₂ by R ₁ in some cases and the like. For example, based on the performance of class 1, class 2 sets R ₁ = 1 and R ₂ = 1.5 when aiming for 1.5 times performance.

1 is a configuration diagram of an IP-SAN to which the class-based TCP congestion control method of the present invention is applied.

The terminal 110 of the SAN user using iSCSI in the home or enterprise is connected to the SAN 130 through the Internet network 120.

Referring to FIG. 1, an embodiment in which the congestion control method of the present invention is applied to an IP-SAN will be described.

Since non-real-time data traffic such as web data or non-real-time VOD is not required for real-time performance, it is set to Δ ₁ = 4. Then, since the relative performance R can be calculated using Equation 2, it is expressed as Equation 5 below.

Since Δ ₂ satisfying the desired relative performance R can be calculated using Equation 3, it is expressed as Equation 6 below.

Table 1 is a comparison table comparing the relative performance R and Δ ₂ when Δ ₁ = 4.

Δ ₂ R 4/3 2.05 3/2 1.73 2 1.34 3 1.10 4 1.00 6 0.92

When Δ ₁ = 4/3 and Δ ₂ = 4 are used to double the relative performance between the two classes, Δ A component ratio for each class for satisfying the relative performance R _ref using = 2 may be calculated using Equation 4, and thus is expressed as Equation 7 below.

Referring to Table 1, when R ₂ of Δ ₂ = 4, that is, R ₁ = 2.05 to achieve 2.05 times the performance when R ₂ = 1, Δ ₁ = 4/3 Can be obtained, Δ If you follow the conventional method of using = 2, the value is R _ref = 1.34. Substituting these values into Equation 4 results in Equation 7.

That is, in order to provide 2: 1 relative performance for the class X of the real-time traffic characteristic and the class Y of the non-real time characteristic, and to maintain the characteristics of the existing network, the reduction factors Δ ₁ = 4 and Δ ₂ = 4/3 are used. The ratio of traffic between the two classes is 1: 2. When the composition ratio of traffic is 1: 2, Δ The real-time service of Class 1 is 1 when two non-real-time services are provided to two users to satisfy the desired performance ratio of 2: 1 while maintaining the same characteristics as the existing network following the conventional method of using = 2. It should only be given to people.

The present invention can be applied to a domain or a network unit according to a user's request, and it is possible to provide a requested service when the applied network is an end-to-end exclusive single path. And, if it is not an end-to-end exclusive single path, the standard (Δ) is calculated using Equation 7 calculated with reference to Equation 4. = 2) can be applied while maintaining traffic and equality conditions.

In other words, if the Δ is adjusted to improve the performance of the network, the normal network becomes difficult to use because it affects the performance of other normal traffic. It does not affect the performance of existing normal traffic of = 2. The present invention can also be applied to an HBA (Host Bus Adapter) for an IP-SAN storage server.

2 is a flowchart of a TCP congestion control method through TCP data transmission window adjustment according to the present invention.

First, the performance of the real-time service class network in which the TCP data of the real-time service is transmitted and the performance of the non-real-time service class network in which the TCP data of the non-real time service is transmitted are calculated (S210). The performance of the network is calculated using Equation 1 described above.

Then, the relative performance is calculated using the performance information of the real-time service class network and the performance information of the non-real-time service class network (S220). The calculation of the relative performance is made using Equation 2. The relative performance equation can be summarized to obtain the transmission window reduction factor. That is, the transmission window reduction factor of the real-time service class network or the transmission window reduction factor of the non-real-time service class network is calculated using the relative performance information (S230).

Using the calculated transmission window reduction factor, congestion of TCP traffic is controlled by adjusting the size of the TCP data transmission window according to the type of TCP data to be transmitted (S240).

The invention can also be embodied as computer readable code on a computer readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data that can be read by a computer system is stored. Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage, and the like, and may also be implemented in the form of a carrier wave (for example, transmission over the Internet). Include. The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion.

So far I looked at the center of the preferred embodiment for the present invention. Those skilled in the art will appreciate that the present invention can be implemented in a modified form without departing from the essential features of the present invention. Therefore, the disclosed embodiments should be considered in descriptive sense only and not for purposes of limitation. The scope of the present invention is shown in the claims rather than the foregoing description, and all differences within the scope will be construed as being included in the present invention.

As described above, the present invention can provide a differential service according to the characteristics of the Internet data and the user's request without modifying or replacing the existing network intermediate node equipment by using the TCP congestion control algorithm of the transmitter. Therefore, there is an effect that can provide a differentiation function for the existing equipment difficult to separate processing for the identity bit rate service.

1 is a configuration diagram of an IP-SAN to which the class-based TCP congestion control method of the present invention is applied.

2 is a flowchart of a TCP congestion control method through TCP data transmission window adjustment according to the present invention.

Claims (7)

Determining whether TCP data to be transmitted to the Internet network is real-time service data or non-real-time service data; And And adjusting the reduction degree of the transmission window size of the TCP data differently according to the type of the service. Setting a relative performance indicative of a ratio between a performance of a real-time service class network in which TCP data of a real-time service is transmitted and a performance of a non-real-time service class network in which TCP data of a non-real-time service is transmitted; Recalculating a transmission window reduction factor of the real-time service class network or a transmission window reduction factor of the non-real-time service class network according to the set relative performance value; And And adjusting the TCP data transmission window size using the calculated transmission window reduction factor according to the type of TCP data to be transmitted. The method of claim 2, Δ is the reduction factor of the transmission window size, RTT is the round trip time, and b is When acknowledgment signal (ACK) is sent after receiving several packets, and p means packet loss rate, the performance of the real-time service class network and the performance of the non-real-time service class network are as follows. A method for controlling congestion of TCP traffic, characterized in that The method of claim 2, When Δ ₁ is a transmission window reduction factor for the real-time service, and Δ ₂ means a transmission window reduction factor for the non-real-time service, the relative performance is calculated by the following equation. To control congestion. The method of claim 2, When R means relative performance and Δ ₁ means transmission window reduction factor for the real-time service, the transmission window reduction factor Δ ₂ for the non-real-time service is calculated by the following equation. How to control the congestion of traffic. The method of claim 3, X denotes the class of the real-time traffic characteristic, Y denotes the class of the non-real-time characteristic, and R _ref represents the class of the real-time traffic characteristic and the class of the non-real-time characteristic when the reduction factor Δ = 2 of the transmission window size. Relative performance ratio, R ₁ , When R ₂ denotes the relative performance ratio of the real-time traffic characteristic class and the non-real-time characteristic class, respectively, after performing the step of adjusting the TCP data transmission window size, the composition ratio of traffic for the type of service is The method of controlling congestion of TCP traffic further comprising the step of calculating by an equation. X: Y = (R _ref -R ₂ ) :( R ₁ -R _ref ) A computer-readable recording medium having recorded thereon a program for executing the method of claim 2 on a computer.