KR20060111323A - Apparatus and method for controlling data transmission in tcp - Google Patents

Abstract

An apparatus and a method for controlling TCP-based communication are provided to shorten the time required for transmitting/receiving the overall data and considerably improve transmission efficiency. A packet loss detecting unit(420) detects whether a packet transmitted from a transmission side has been lost due to an error at a receiving side. When a packet has been lost due to an error of the receiving side, a packet loss signal transmission unit(430) transmits a response signal indicating the packet loss. As the response signal, a plurality of same ACKs are continuously transmitted. The packet loss detecting unit(420) detects the error of the receiving side from state information of an Ethernet controller(400) or a received packet.

Description

TCCP-based communication control device and method {Apparatus and method for controlling data transmission in TCP}

1 is a flowchart illustrating a data transmission algorithm in a general TCP-based communication system.

2 is a block diagram showing the configuration of a general network system.

3 is a block diagram illustrating a detailed configuration of a network communication module of FIG. 2.

4 is a block diagram showing the configuration of a TCP-based communication control apparatus according to an embodiment of the present invention.

5 is a flowchart illustrating a TCP-based communication control method according to an embodiment of the present invention.

6a and 6b compare the performance between the general communication control method and the communication control method according to the present invention.

The present invention relates to a communication system based on a transmission control protocol (hereinafter referred to as TCP), and more particularly to a system for adjusting a data transmission speed by adjusting a window size and waiting for a backoff time according to a response of a receiver. The present invention relates to a TCP-based communication control apparatus and method for efficient transmission.

TCP is used to transmit data in the form of packets between nodes through the Internet, and adjusts the transmission rate in consideration of network conditions.

In a TCP-based communication system, a transmitting side converts packet data to be transmitted into data segments of a predetermined format and provides the receiving side in window size units. Here, the data segments include sequence numbers for ordering, and the window size is determined by a smaller value between a congestion window (hereinafter referred to as CWND) and a notification window (hereinafter referred to as ADWND). . Here, CWND is a value for estimating the transmission capacity of the network and ADWND is a value representing the reception capability of the receiver, and represents the current buffer size remaining on the receiver.

The receiving end transmits an acknowledgment signal (hereinafter referred to as ACK) when the data segment is normally received. The ACK contains the sequence number of the next data segment normally received.

If the sender does not receive an ACK within the preset timeout period, the network assumes that congestion has occurred on the network and reduces the CWND to one segment size. Will be

However, the cause of not receiving the ACK, that is, the packet loss, may occur not only in congestion on the network but also in the receiving side such as a failure of the bus occupied by DMA (Direct Memory Access). In this case, since no congestion has occurred, it is unnecessary to reduce the transmission rate such as reducing the CWND to one segment size and waiting for the exponential backoff time.

Therefore, in a general TCP-based communication system, even if the cause of packet loss is caused by a problem on the receiver side rather than network congestion, the window size is reduced because the sender does not receive an ACK. Waiting for a while results in inefficient transmission.

The technical problem to be solved by the present invention is that in a TCP-based communication system, when a packet is lost due to a problem on the receiving side due to a resource shortage such as BUS, memory, CPU, etc. It is to provide a TCP-based communication control device and method that can greatly improve the transmission efficiency, such as by reducing the time required to transmit and receive the entire data by transmitting data packets as much as the previous window size.

TCP-based communication control apparatus according to the present invention for achieving the above technical problem is a transmission control protocol (TCP) based communication system including an Ethernet control unit, detecting whether a packet transmitted from a transmitting side is lost due to an error on the receiving side A packet loss detector; And a packet loss signal transmitter configured to transmit a response signal indicating the packet loss when the packet is lost due to the reception side error.

In the apparatus, the response signal preferably transmits a plurality of identical ACKs. In the apparatus, preferably, the packet loss detection unit detects the reception error from the state information of the Ethernet control unit or the received packet.

The TCP-based communication control method according to the present invention for achieving the above technical problem is a transmission control protocol (TCP) based communication system including an Ethernet control unit, detecting whether the packet transmitted from the transmitting side is lost due to the error of the receiving side Doing; And transmitting the response signal indicating the packet loss when the packet is lost due to the error in the receiving side.

In the above method, it is preferable that the response signal continuously transmits a plurality of identical ACKs.

Hereinafter, a method and an apparatus according to the present invention will be described in detail with reference to the accompanying drawings.

1 is a flowchart illustrating a data transmission algorithm in a general TCP-based communication system.

First, the transmitting side and the receiving side make a TCP connection in a three way handshake (step 100).

If a TCP connection is established between the transmitting side and the receiving side, the transmitting side converts the packets into the data segment format and transmits the data segments by the window size (step 105). After transmitting the data packet, the receiver waits for an ACK sent by the receiver with a preset timeout time (step 110).

Next, it is determined whether an ACK has been received during the timeout period (step 115), and it is assumed that network congestion has occurred, and the slow start threshold (hereinafter referred to as SSTHRESH) is half of the current window size. Reduce the CWND to one segment size, wait for a predetermined exponential backoff time (step 120) to eliminate congestion, and then return to the data transfer mode (step 105). If CWND is less than the SSTHRESH, it is determined as a slow start mode, otherwise it is determined as a congestion avoidance mode. The difference between the two modes is a method of increasing the CWND when receiving an ACK from the receiving side. The CWND is increased exponentially in the slow start mode and linearly in the congestion avoidance mode.

On the other hand, if the ACK is received within the timeout time, it is determined whether the ACK has the expected sequence number (step 125).

If the received ACK is not the expected ACK, it is determined whether it is a triple duplicate ack (step 130). It is determined whether an ACK having the same sequence number as the sequence number of the currently received ACK has been received twice before. Duplicate ACK is a signal generated by the receiver when the received segments are out of order and used for fast retransmission. Fast retransmission is a transmission mode in which a segment having a corresponding sequence number is considered lost when a predetermined number of duplicate ACKs are received, and immediately retransmits the missed segment. That is, when receiving consecutive duplicate ACKs of more than the Retransmit Threshold, the sender immediately retransmits the segment regardless of whether the retransmission timer expires. Normally, the retransmission threshold is three.

If it is determined in step 130 that the received ACK is not a triple redundant ACK, the process returns to the waiting step (step 110) for waiting for the ACK before timeout.

On the other hand, if it is determined in step 130 that the received ACK is a triple redundant ACK, SSTHRESH is set to half of the current window size, and CWND is set to be 3 segments larger than SSTHRESH (step 135). That is, unlike the step 120, if a triple redundant ACK is received, the CWND is not significantly reduced or the exponential backoff time is waited. That is, even if the retransmission timer does not expire, the process proceeds to the data transmission step 105 and retransmits the segment having the sequence number included in the triple redundant ACK.

If it is determined in step 125 that the received ACK is the expected ACK, it is determined whether all data transmission is completed (step 140).

If there is data left to be transmitted, the CWND is incremented (step 145) and the data returns to step 105 (step 105). In this case, as described above, the CWND is increased exponentially in the slow start mode and linearly in the congestion avoidance mode.

If there is no data to transmit, the TCP connection is disconnected (step 150), and the data transmission algorithm is terminated.

FIG. 2 is a block diagram illustrating a general network system including a CPU core 200, a network communication module 210, an SDRAM controller 220, and other units 230 and 240.

The CPU core 200 is responsible for arithmetic processing, and the network communication module 210 transmits and receives packets using a LAN. The SDRAM controller 220 is in charge of controlling the SDRAM.

3 is a block diagram illustrating a detailed configuration of the network communication module 210 of FIG. 2, which includes a physical layer processor 300 and an Ethernet controller 310, and the Ethernet controller 310 is a MAC. (Media Access Control) core 320, DMA (Direct Memory Access) 330 is included.

The operation of each block is described below in terms of data reception.

The physical layer processor 300 is in charge of a physical layer and is connected to a wired or wireless LAN. That is, the physical layer processing unit 300 establishes a link to the counterpart network system or the relay device according to the set value, and when the link is established, receives the packet from the network and sets the MAC core in the Ethernet control unit 310. Forward to 320.

In the Ethernet control unit 310, the MAC core 320 receives the Ethernet packet from the physical layer processing unit 300 and checks whether there is an abnormality of the packet, and determines whether to receive and receive a destination address. If it is determined to receive and store, the received packet is delivered to the DMA 330. The DMA 330 occupies a bus to store the packet in SDRAM and to generate an interrupt. On the other hand, the result of the inspection of the abnormality of the packet and the status information such as whether the DMA 330 has failed to occupy the bus and whether the received packet is lost because there is no space to store in the SDRAM, the status information in the Ethernet controller 310 Stored in a storage location, the state information storage location is generally implemented as a register or the like.

The CPU core 200 receiving the interrupt from the network communication module 210 processes the packet data stored in the SDRAM by the DMA 330 in the order of a data-link layer, an IP layer, and a TCP layer. The TCP layer transmits an ACK for the received packet to the sender.

On the other hand, when the sender does not receive the ACK for the transmitted data, it is determined that congestion has occurred on the network, and thus the lost data is lost. As shown in step 120 of FIG. 1, the sender reduces the window size and stops the transmission during the exponential backoff time. To slow down. However, even though there is no congestion on the network, a situation may occur in which data is lost due to a problem at the receiving end. An example of such a situation is shown below.

1. DMA 330 fails to occupy the bus and loses data-DMA 330 fails to occupy the bus due to high bus usage during certain tasks in the system in which the network communication module 210 is embedded. can do.

2. When DMA 330 loses data because there is no space to accumulate data in SDRAM-While using a specific operation in the system in which the network communication module 210 is embedded, the CPU usage increases and the DMA is stored in the SDRAM. You can't process the data you've placed quickly, and you won't be able to free up storage space, which can result in data loss.

3. Other MAC cores 320 may encounter an error such as an alignment error or a cyclic redundancy check (CRC) error, thereby losing data.

Therefore, if there is a packet loss due to the problem of the receiver as described above, unlike the conventional retransmission processing method, a predetermined response signal is sent to the transmitter so that the transmitter can determine that the packet is not lost due to network congestion. As a result of this determination, it is a principle of the present invention to prevent unnecessary transmission speed reduction.

4 is a block diagram showing the configuration of the TCP-based communication control apparatus according to an embodiment of the present invention, the Ethernet control unit 400, the receiving packet storage unit 410, packet loss detection unit 420, packet loss signal transmission unit 430.

The Ethernet controller 400 checks whether there is an abnormality of the received Ethernet packet and a destination address and determines whether to receive the received Ethernet packet, and then provides the packet to a received packet storage unit 410 such as an SDRAM through DMA in the case of a receivable packet. If it is not a receivable packet, the packet is discarded. The abnormality of the received packet may be performed through a CRC check. The CRC check may be performed by the Ethernet controller 400 before storing in the received packet storage 410, and once the received packet is stored in the received packet storage 410, the CRC check may be called again if necessary. You can also do On the other hand, the status information such as the result of the inspection of the abnormality of the packet and whether the DMA has failed to occupy the bus, or whether the received packet is lost because there is no space to store in the SDRAM or the like is stored in the Ethernet control unit 400. It is stored in a place, the storage place is generally implemented as a register or the like.

The received packet storage unit 410 stores the received packet. Here, when the CRC check is performed by the Ethernet controller, it is not necessary to transfer the stored packet to the packet loss detector 420. However, when the CRC check is performed by the packet loss detector 420 instead of the Ethernet controller 400, the stored packet is stored. The packet is delivered to the packet loss detector 420 for CRC checking.

The packet loss detecting unit 420 determines whether the packet is lost due to a problem of the receiving side through the state information of the Ethernet control unit 400 or through a CRC check of the packet stored in the receiving packet storage unit 410. The result is transmitted to the packet loss signal transmitter 430.

When the packet loss signal transmitter 430 receives that the packet is lost due to a problem on the receiver side, the packet loss signal transmitter 430 sends a predetermined response signal indicating that the packet is lost due to the problem on the receiver side. Here, the predetermined response signal is preferably a predetermined number of duplicate ACKs (duplicate ack) to be compatible with the general transmission algorithm as shown in Figure 1, among which is a triple duplicate ACK commonly used for fast retransmission desirable.

When the transmission side receives the predetermined response signal from the reception side, the transmission side goes through step 135 instead of step 120 of FIG. 1, thereby preventing unnecessary transmission speed reduction.

5 is a flowchart illustrating a TCP-based communication control method according to an embodiment of the present invention.

The Ethernet control unit 400 stores the received packet in the received packet storage unit 410 when it is determined that there is no abnormality after checking whether there is an abnormality of the received packet, and discards the received packet if it is determined that there is an error (500). step). The abnormality of the received packet can be determined by determining whether the destination address is correct or by checking the CRC. The CRC check may be performed by the Ethernet controller 400 or the packet loss detector 420. When the CRC check is performed by the Ethernet controller 400, the result of the check is stored in a status information register in the Ethernet controller 400. . The storage of the received packet storage unit 410 may be performed by storing the data in the SDRAM by occupying the bus in the DMA 300. In this case, when the DMA 330 fails to occupy the bus and loses data, or when the DMA 330 loses data because there is no space to store data in the SDRAM, such information may also be generated. In the status information register in 400).

The packet loss detector 420 checks the state information in the Ethernet controller 400 (step 510). When the CRC check is performed by the Ethernet control unit 400, the CRC check result is included in the status information. When the CRC check is performed by the packet loss detecting unit 420, the CRC check is performed by loading the stored received packet. Perform.

The packet loss detecting unit 420 determines whether a packet is lost from the state information (step 520).

If it is determined in step 520 that the packet is lost, the packet loss signal transmitter 530 transmits a triple redundant ACK to the transmitter (step 530). Thereafter, the transmitting side transmits the lost packet in a fast retransmission mode upon receiving the ripple duplicate ACK. That is, instead of performing step 135 of FIG. 1 which reduces the window size and waits for the exponential backoff time, step 120 is performed. As a result, unnecessary transmission speed is not reduced, thereby enabling efficient transmission.

If it is determined in step 520 that the packet is not lost, normal packet processing is performed (step 540). That is, the received packet goes through the TCP / IP layer step through the data link layer.

6A and 6B compare the performances of the general communication control method and the communication control method according to the present invention when a packet is lost due to a problem on the receiving side.

6A shows the operation of the general communication control method when a packet is lost due to a problem on the receiving side.

The transmitting side transmits the first to sixth segments corresponding to the current window size. On the other hand, if a packet is lost due to a bus occupancy failure or the like in the first to sixth segments, the receiving side does not send an ACK by treating it like a normal packet loss. Since the sender did not receive the ACK for the first packet within the timeout time, the transmitter proceeds to step 120 of FIG. 1 and waits for the exponential backoff time to reduce the window size to 1 to send only the first segment. When the next sender receives the ACK for the first segment, the window size is increased by two segments to transmit the second and third segments.

6B illustrates the operation of a general communication control method in the case where a packet is lost due to a problem on the receiving side.

The transmitting side transmits the first to sixth segments corresponding to the current window size. On the other hand, when a packet is lost due to a bus occupancy failure or the like in the first to sixth segments, the receiver sends a triple redundant ACK to the transmitter. The transmitting side retransmits the first to sixth segments at the six-segment size, which is the previous window size, without waiting for the exponential backoff time upon receiving the triple overlap ACK. As a result, according to the present invention, the transmission time is shortened by the exponential backoff time, and since there is no unnecessary reduction of the window size, the transmission time is also reduced.

Such a method and apparatus of the present invention have been described with reference to the embodiments shown in the drawings for clarity, but these are merely exemplary, and various modifications and equivalent other embodiments are possible to those skilled in the art. Will understand. Therefore, the true technical protection scope of the present invention will be defined by the appended claims.

In the TCP-based communication system, when a packet is lost due to a problem on the receiving side due to lack of resources such as BUS, memory, CPU, etc., the data of the previous window size is not waited for by the transmitting side. By transmitting the packet, the transmission efficiency can be greatly improved, such as reducing the time required for transmitting and receiving the entire data.

Claims (5)

A transmission control protocol (TCP) based communication system comprising an Ethernet control unit, A packet loss detector for detecting whether a packet transmitted from the transmitting side is lost due to an error of the receiving side; And And a packet loss signal transmission unit for transmitting a response signal indicating the packet loss when the packet is lost due to the reception side error. The method of claim 1, And the response signal transmits a plurality of identical ACKs in series. The method according to claim 1 or 2, The packet loss detecting unit detects the reception side error from the state information of the Ethernet control unit or the received packet. A transmission control protocol (TCP) based communication system comprising an Ethernet control unit, Detecting whether a packet transmitted from a transmitting side is lost due to an error at the receiving side; And And transmitting a response signal indicating the packet loss when the packet is lost due to the reception error. The method of claim 4, wherein The response signal is a communication control method, characterized in that for transmitting a plurality of the same ACK consecutively.

Images