KR20060056810A - Apparatus and method for using separation tcp buffer performance degradation protection tcp flow - Google Patents

Abstract

Apparatus and method for preventing TCP flow degradation using TCP buffer separation according to the present invention include a class filter unit for classifying TCP packets and non-TCP packets by checking a protocol of an input packet and transmitting the classified and non-TCP packets; And a buffer unit including at least one or more multiple buffers each receiving the same packet from the class filter unit and buffering the packet according to each protocol, thereby preventing a sudden performance degradation of the TCP flow in a buffer overflow situation.

Description

Apparatus and method for using separation TCP buffer performance degradation protection TCP flow}             

1 is a diagram illustrating flow control of a packet of a general TCP flow;

2 is a diagram illustrating flow control of a packet of a general UDP flow;

3 is a diagram illustrating a state of processing a packet of a buffer according to FIGS. 1 and 2;

4 is a functional block diagram showing a TCP flow performance degradation prevention apparatus using TCP buffer separation according to an embodiment of the present invention;

5 is a view showing a state of the buffer unit of the TCP flow performance degradation prevention apparatus using the TCP buffer separation according to FIG.

6 is a flowchart illustrating a method for preventing TCP flow degradation by using TCP buffer separation according to an embodiment of the present invention.

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

100: class filter unit 200: buffer unit

210: TCP Buffer 220: Non TCP Buffer

The present invention relates to an apparatus and method for preventing TCP performance degradation using TCP (Transmission Control Protocol) buffer separation. More particularly, the present invention relates to a buffer that separates a buffer of TCP that performs flow control in an environment of the Internet. The present invention relates to an apparatus and method for preventing TCP performance degradation by using TCP buffer separation to improve the performance of TCP flow in a buffer overflow situation.

Conventional buffer management method performs buffer management without distinguishing the protocol of the upper IP in one buffer. In other words, the packet discard in the buffer is caused by the overflow of the output buffer when a larger amount of packets than the capacity of the output put (Out Put) of the output link has to be transmitted in the direction of the output link. Packet dropping is performed using a tail drop method. The tail drop is a queue in the form of First In First Out (FIFO) / First In First Service (FIFS). Discard the last packet that came in. In this case, the hardware can be simply implemented, but the data rate is different for each flow. Bad nodes send their own packets. However, Drop Tail Queue discards incoming packets at the end, no matter whether they are bad or good, so you can't prevent bad nodes from monopolizing network resources. Several methods have been proposed to solve this. Among them, Nagle's proposed method is to distribute each flow fairly by the router. Look at the source address of the packet and have a separate queue for each flow. Therefore, even if a bad node sends out a packet, it is only discarded in its own queue and does not interfere with the good node. In addition, FQ, SFR, and DRR came out to create a fair flow.)

By default, as traffic increases in the router's buffer, the buffer begins to fill up with packets. If traffic continues to pass through the router at a rate that exceeds the output link speed, it can fill up to the maximum capacity of the buffer. In this case, the router has to discard all subsequent packets until there is enough buffer, and randomly occurs when the packets are not classified.

As described above, the characteristics of TCP, which is a representative protocol for performing flow control, and UDP, which is a representative protocol for not performing flow control, will be described.

First, Transmission Control Protocol / Internet Protocol (TCP / IP) is the basic communication protocol of the Internet, and is also used in private networks such as intranets and extranets. When a user sets up his computer to access the Internet, a TCP / IP program is installed, which allows him or her to exchange messages or obtain information with other computer users who are also using the same TCP / IP protocol.                         

TCP / IP is a two-layer program. The upper layer, TCP, is responsible for breaking messages or files into smaller packets for transmission over the Internet, and for reassembling the received packets into original messages. The lower layer, or IP, processes the address portion of each packet, allowing packets to reach their destination exactly. Each gateway on the network checks the address of the message to know where to send the message. If a message is divided into packets, each packet can be delivered through a different path, and they are reassembled at their final destination.

TCP / IP uses a client / server model for communication, in response to the needs of computer users (clients), providing other services on the network (servers) to send web pages. TCP / IP inherently makes point-to-point communication, which means that each communication starts from one point (or host computer) on the network and is forwarded to another point or host computer. Both TCP / IP and higher-level applications that use TCP / IP are called "connectionless", meaning that each client's request is considered a new request, independent of any previous request. (Different from having a regular phone call connected continuously throughout the call). Since connectionless does not monopolize the network, it allows everyone to use the path constantly in common (note that in fact the TCP layer itself is not connectionless as long as a message is involved. Until all packets belonging to a message are received).

Many Internet users are familiar with higher layer application protocols using TCP / IP. These higher-level protocols include HTTP for Web services, Telnet for logging on to remote computers, and FTP for file transfer and SMTP for mail. These protocols are often packaged together with TCP / IP.

PC users usually use the SLIP or PPP protocols to access the Internet. These protocols encapsulate IP packets so that they can be sent to the service provider's modem via dial-up dial-up.

A protocol related to TCP / IP is UDP, which is used instead of TCP for special purposes. Protocols used by network host computers to exchange routing information include ICMP, IGP, EGP, and BGP.

In addition, UDP (User Datagram Protocol) [UDP] is a communication protocol that provides only limited services when messages are exchanged between computers in a network using IP. UDP is an alternative to TCP and sometimes referred to as UDP / IP when used with IP. Like TCP, UDP uses IP to receive the actual data units called datagrams from one computer to another. However, unlike TCP, UDP does not provide services such as dividing a message into packets (datagrams), reassembling on the other side, and in particular does not provide an order of arriving data packets. This means that applications using UDP must be able to verify that the entire message has arrived in the correct order. Network applications that have very little data to exchange (and therefore very few messages to reassemble) may prefer UDP over TCP to reduce processing time. TFTP, for example, uses UDP instead of TCP.

UDP provides two services that are not provided in the IP layer. One is a port number for distinguishing other user requests, and a checksum function for checking whether the data arrives corrupted.

In the OSI communication model, UDP belongs to the transport layer, which is four layers like TCP.

However, most data packets transmitted / received in the conventional Internet have been transmitted using TCP and UDP.

Referring to the flow control by packet discard of the conventional TCP flow, as shown in FIG. 1 showing the change in the TCP transmission rate during TCP packet transmission, the average transmission rate of the TCP flow is calculated as follows.

Average rate of TCP flows =

B: maximum transfer rate

E: The transmission rate exceeding the maximum transmission rate (this part is discarded by buffer management). E is the size at which packet discard occurs and is actually a very small value.

As a result of the calculation, the TCP flow has flow control due to packet discard in a buffer overflow situation. That is, as shown in FIG. 1, the TCP packet existing between B + E and B is discarded and the TCP flow dramatically lowers the transmission rate.

In the router's buffer, packets are not classified and stored in a buffer. When an overflow occurs in the buffer, the packet is discarded randomly from the buffer. To control this, the TCP flow dramatically lowers the transmission rate. Therefore, the TCP flow had a problem that the transmission rate is too low due to flow control when an overflow situation occurs.

2 illustrates an operation in which a packet discard situation occurs in UDP, which is a protocol that does not perform conventional flow control.

As shown in FIG. 2, even if packet discard occurs, flows not controlled by the flow control cannot deliver only discarded packets, and the remaining packets are normally delivered.

As shown in FIG. 2, only non-TCP packets existing between B + E and B are discarded while the remaining non-TCP packets are transmitted as they are. Therefore, flows without flow control are not greatly affected by the transmission rate due to packet discard.

Therefore, as shown in FIGS. 1 and 2, when the flow control traffic and the traffic without traffic are delivered together, the flow control traffic loses more than the traffic traffic without the traffic control.

If a buffer overflow of a router occurs in a situation where multiple TCP flows and UDP flows exist together, multiple TCP flows may cause packet discard at the same time, and each TCP flow may be retransmitted from the receiving side to the transmitting side. On the request side, the sending side dramatically reduces the transmission rate of the transmitting TCP.

On the contrary, since UDP flows do not use flow control, all packets except UDP packet transmission rate are transmitted, but TCP transmission rate is lowered.

Accordingly, the present invention has been made to solve the above-mentioned conventional problems, and an object of the present invention is to prevent the TCP rate of flow control from falling sharply because the TCP buffer is separated and managed in a buffer overflow situation. The present invention provides an apparatus and method for preventing TCP performance degradation using TCP buffer separation.

According to an aspect of the present invention, there is provided a TCP flow degradation prevention apparatus using a TCP buffer separation. The apparatus includes: a class filter unit configured to classify TCP packets and non-TCP packets by checking a protocol of an input packet and then transmit the classified packets; And a buffer unit for receiving the same packet from the class filter unit and buffering the packet according to each protocol.

On the other hand, according to one aspect of the TCP flow performance degradation prevention method using the TCP buffer separation according to the present invention, the first step of classifying the TCP packet and Non TCP packet by the class filter unit to check the protocol of the input packet; And a second step in which the buffer unit receives the same packet from the class filter unit and performs buffering corresponding to each packet.

Hereinafter, an apparatus and method for preventing TCP flow performance degradation using TCP buffer separation according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

4 is a functional block diagram of a TCP flow performance degradation prevention apparatus using TCP buffer separation according to an embodiment of the present invention, TCP flow performance degradation prevention apparatus using TCP buffer separation according to an embodiment of the present invention is a class filter The unit 100 and the buffer unit 200 are included.

When the class filter unit 100 simultaneously inputs a TCP packet and a non-TCP packet, the class filter unit classifies the protocol of the packet, classifies the TCP packet and the non-TCP packet, and transmits the packet to the buffer unit 200.

The buffer unit 200 receives the same packet from the class filter unit 100 and buffers the packet corresponding to each protocol. At this time, the buffer unit 200 includes a TCP buffer 210 for buffering TCP packets, and a non-TCP buffer 220 for buffering packets except TCP packets.

In addition, the scheduler receives a TCP packet and a Non TCP packet output from the buffer unit 200 and processes and outputs each protocol.

Next, an operation process of the TCP flow performance degradation prevention method using the TCP buffer separation having the above configuration will be described with reference to FIG. 6.

First, the class filter unit 100 checks the protocol of the input packet and classifies the TCP packet and the non-TCP packet and transmits them (S1). That is, as shown in FIG. 3, when the TCP packet and the Non TCP packet are input, the class filter unit 100 may classify the IP packet by checking the IP of the TCP packet and the Non TCP packet.

Subsequently, the buffer unit 200 receives and buffers the same packet from the class filter unit 100 (S2). That is, the buffer unit 200 performs buffering through the TCP buffer 210 when receiving the TCP packet from the class filter unit 100 as shown in FIG. 5, and non-TCP when receiving the non TCP packet. Buffering is performed through the buffer 220.

Therefore, the TCP buffer flow control is performed through the TCP buffer 210, and the non-TCP buffer flow control is performed through the non-TCP buffer 220 and after discarding the overflowed packet. The rest of the packet is sent as is.

Although the present invention has been described in detail only with respect to the described embodiments, it will be apparent to those skilled in the art that various modifications and changes are possible within the technical spirit of the present invention, and such modifications and modifications belong to the appended claims.

As described above, according to the apparatus and method for preventing TCP flow degradation using TCP buffer separation according to the present invention, there is an excellent effect of preventing a sudden degradation of TCP flow in a buffer overflow situation.

Claims (6)

A class filter unit for classifying a protocol of an input packet and classifying a Transmission Control Protocol (TCP) packet and a Non TCP packet; And And a buffer unit for receiving the same packet classified from the class filter unit and buffering the packet according to each protocol. The method of claim 1, And the buffer unit includes at least one TCP buffer for buffering TCP packets and at least one non-TCP buffer for buffering packets other than TCP packets. The method of claim 1, Device for preventing TCP performance degradation using TCP buffer separation comprising a scheduler for scheduling and transmitting the buffered TCP packets and Non TCP packets from the buffer unit. A first step of classifying the packet of the input packet by the class filter unit, classifying the TCP packet and the non-TCP packet and transmitting the classified packet; And And a second step of the buffer unit receiving the same packet from the class filter unit and performing buffering corresponding to each packet. The method of claim 4, wherein The second step, And a twenty-first step of receiving a TCP packet from the class filter and performing a flow control of the TCP packet by the TCP buffer of the buffer unit. The method of claim 4, wherein The second step, And a twenty-second step in which the non-TCP buffer receives non TCP packets from the class filter and does not perform flow control of the non TCP packets.

Images