WO2015074279A1

WO2015074279A1 - Network encoding and transmission method based on udp protocol

Info

Publication number: WO2015074279A1
Application number: PCT/CN2013/087789
Authority: WO
Inventors: 李挥; 谭学磊; 秦立都; 于超琪; 潘凯
Original assignee: 北京大学深圳研究生院; 深圳赛思鹏科技发展有限公司
Priority date: 2013-11-25
Filing date: 2013-11-25
Publication date: 2015-05-28
Also published as: CN107852398A

Abstract

The present invention relates to a network encoding and transmission method based on the UDP protocol, comprising the following steps: establishing a network encoding layer that is connected to a transmission layer by using the UDP protocol, wherein data or a file enters from an application layer of a transmit end to a network encoding layer of the transmit end, and is processed in the network encoding layer of the transmit end to form network code, a data packet with a number is generated in a set manner by using the network code, and the data packet is sent to a receive end by the transmission layer of the transmit end; a network encoding layer of the receive end receiving the encoded data packet by a transmission layer of the receive end, identifying the number of the received data packet, returning a different response signal to the transmit end according to the received number, decoding the data packet to obtain data or a file, and transmitting the data or the file to an application layer of the receive end. The network encoding and transmission method based on the UDP protocol has the following beneficial effects: it is easy to implement the method, a receive end has a short delay, and the method adapts to the change in network environment.

Description

Network coding and transmission method based on UDP protocol

Technical field

The present invention relates to the field of network coding and transmission, and more particularly to a network coding and transmission method based on the UDP protocol.

Background technique

At present, the research of network coding in the transport layer is mainly based on the TCP protocol, such as the TCP protocol based on network coding theory (TCP/NC). In the prior art, in order to combine the advantages of TCP and network coding, a new protocol called TCP/NC is proposed. The TCP/NC packet uses a random linear network coding to encode a single data stream, modifying the TCP acknowledgment (ACK) scheme so that it responds to the acknowledgment of freedom (the number of new encoded packets received) instead of a single packet. TCP/NC uses random linear coding techniques to mask the loss of TCP packets and improve throughput. However, for the increasingly widely used wireless networks, the application of the TCP protocol is not good. This is because the traditional transport layer protocol will treat packet loss as network congestion, which will reduce the congestion window and lead to a decrease in network throughput. At present, the TCP protocol method based on network coding theory (TCP/NC), that is, adding a network coding layer (NC layer) between the TCP layer and the IP layer, requires some modifications to the existing TCP protocol. The NC layer at the transmitting end receives the data packets from the TCP layer and stores them in an encoding buffer representing the encoding window, encodes the buffered original data packets, and then repackages the encoded data packets and transmits them; The layer is responsible for receiving the encoded data packet from the network and buffering, then returning an ACK message to the transmitting end, decoding the received encoded data packet, and finally transmitting the decoded original data packet to the TCP layer. The sender generates and transmits a random linear combination of data packets in the coding window, and the linear combination coefficients are transmitted through the packet header.

In this case, for the data packets in the coding window, R random linear combinations will be performed (R is the redundancy coefficient, which is the reciprocal of the successful reception probability, which may not be an integer, so it will be rounded up during the operation) Go out. When the network condition is relatively stable and the packet loss rate does not change much, the network throughput and link utilization can be improved by setting a fixed optimal redundancy coefficient R (the optimal redundancy coefficient R is equal to the reciprocal of the successful reception rate). rate. However, in an environment where the network situation is unknown and the packet loss rate changes frequently, the R of the redundancy coefficient is not easy to determine. If R is too small, it will not be enough to cover the loss of the data packet. Therefore, a large number of timeouts may occur, resulting in low throughput; Conversely, the transmission rate is limited by the rate of the encoding itself, which also reduces throughput, and sending too many linear combinations can also block the network. Therefore, in the prior art, in addition to the modification of the TCP protocol is large and difficult to implement, its encoding using the coding window may also bring a large delay at the receiving end, and at the same time fix the value of the redundant coefficient.

The TCP/NC protocol cannot adapt to the changing environment of the network environment, especially the wireless network environment that is greatly affected by external interference.

Summary of the invention

The technical problem to be solved by the present invention is to provide an easy implementation, a small delay at the receiving end, and a change in the network environment, which is difficult to implement in the prior art, has a large delay at the receiving end, and cannot adapt to changes in the network environment. Network coding and transmission method based on UDP protocol.

The technical solution adopted by the present invention to solve the technical problem is: Constructing a network coding and transmission method based on the UDP protocol, comprising the following steps:

A) constructing a network coding layer between the transport layer and the application layer; the network coding layer is respectively connected to its corresponding transport layer and application layer, and the network coding establishes a connection with the transport layer by using the UDP protocol of the transport layer ;

B) The data or file is entered by the application layer of the transmitting end into the network coding layer of the transmitting end, buffered and coded by the network coding layer of the transmitting end, and formed a numbered data packet according to the set manner, through the transmitting end transmission layer, Sending to the receiving end by means of transmission with reliable transmission;

C) The network coding layer of the receiving end receives the encoded data through the transmission layer of the receiving end, identifies the number of the received data packet, returns a different response signal to the transmitting end according to the received number, and decodes the data to obtain the data. Or file, and transferred to the application layer at the receiving end.

Further, the method further includes the following steps:

D) The transmitting end decides to retransmit or not retransmit the data block according to different types of response signals sent by the receiving end.

Further, the step B) further includes the following steps:

B1) when the transmitting end establishes a connection with the receiving end, sending a data packet carrying the set transmission parameter to notify the receiving end, and the receiving end returns the acknowledgement information after receiving the data packet and according to the transmission parameter Set the receiver.

Further, the step B) further includes the following steps:

B2) the data or file to be transmitted by the transmitting end is divided into data packets of a first set length according to the inflow order thereof, and when the remaining file or data length is less than the first set length, the data is filled in. Data zero to maintain the packet length, so that each packet after division has the same length; set the number of packets blks ize to form a data block, the data packet is the smallest unit of transmission; the data contained in the data block The number of packets is adjusted according to the information returned by the receiving end during transmission.

Further, the step B) further includes the following steps:

B3) The transmitting end controls the sending speed of the data packet by the number of tokens allowed to be allowed in a unit time, and each token allows one data packet to be sent; the number of tokens that appear in the unit time is transmitted according to the data packet. Time adjustment.

Further, the number of tokens allowed to be transmitted in a unit time is adjusted according to a transmission time of the data packet at the transmitting end and the receiving end, and the receiving end is notified; the longer the data packet is transmitted, the unit The fewer tokens allowed to be sent during the time.

Further, the number of data packets included in the data block is adjusted according to the transmission time of each data packet, and the number of data packets included in the data block is inversely proportional to the transmission time of the data packet between the transmitting end and the receiving end. .

Further, the step C) further includes the following steps:

C1) the receiving end detects the data packet number that has been received, and if the number is consecutive, sends an acknowledgement response signal to the transmitting end; if the number is not continuous, sends an extended acknowledgement response signal to the transmitting end, the extended acknowledgement signal The last consecutive data packet number received by the receiving end is carried in the middle.

Further, the step C) further includes the following steps:

C2) upon receiving a new data block, initializing a coding coefficient matrix of size blks ize*blks ize and a corresponding payload structure, where blks ize is the length of the currently set data block;

C3) for each received data packet, insert its coding coefficient and payload into the above-mentioned coding coefficient matrix and the corresponding payload structure respectively, and use Gaussian elimination method to judge whether it is independent of the previous data packet, and if so, Perform step C4); otherwise, perform step C5);

C4) returning an acknowledgment signal, updating the current block degree of freedom, adding 1 to it, and determining whether the updated current block degree of freedom is equal to the current set length in the data block, and if so, determining that the data block can be decoded; otherwise, waiting for reception Next packet;

C5) Sends an acknowledgment signal, but does not update the block degrees of freedom and waits to receive the next packet. Further, if the sending end sends a data packet and the acknowledgment signal is not received for longer than the transmission time of the last data packet plus the retransmission timeout period, the transmitting end transmits the data packet in a slow start mode; the slow start mode includes The number of tokens allowed to be transmitted in the unit time is set to a default value, and the number of tokens is incremented by one after each successful transmission of a data packet, until the token number reaches a set threshold, and is changed to a normal transmission state.

The UDP protocol-based network coding and transmission method embodying the present invention has the following beneficial effects: Since the network coding layer is disposed between the application layer and the transport layer, the network coding layer is connected to the transport layer through the UDP protocol stack of the transport layer and The data is transmitted, and the modification difficulty and workload of the UDP protocol are much smaller than the modification of the TCP/IP protocol. At the same time, an acknowledgment signal is used between the receiving end and the transmitting end, and a reliable UDP transmission form is used to ensure reliable transmission of data; The number of data packets transmitted per unit time and the number of data packets included in each data block can be adjusted in real time according to the transmission time of the data packet to adapt to changes in the network environment. Therefore, it is easier to implement, the receiver has less delay, and adapts to changes in the network environment.

DRAWINGS

1 is a schematic diagram of a network coding layer setting in an embodiment of a network coding and transmission method based on a UDP protocol according to the present invention;

Figure 2 is a flow chart showing a case of data transmission in the embodiment;

3 is a flow chart of transmitting data and congestion control at the transmitting end in the embodiment.

Figure 4 is a flow chart of the receiving end decoding in the embodiment;

Figure 5 is a diagram of six state transitions for reliable transmission of the network coding layer. detailed description

The embodiments of the present invention will be further described below in conjunction with the accompanying drawings.

As shown in FIG. 1 and FIG. 2, in the embodiment of the UDP protocol-based network coding and transmission method of the present invention, the method includes the following steps:

Step S11 constructs a network coding layer to be respectively connected to the application layer and the transport layer: In this step, in order to implement coding and data transmission, it is necessary to construct an encoding mechanism, and the mechanism is capable of obtaining the encoded data after encoding. Transmission to the receiving end for decoding; therefore, both the transmitting end and the receiving end need to set a network coding layer, and when the network coding layer is used as a receiving end, the data is compiled. The code obtains the encoded data; and when it is the receiving end, decodes the received encoded data. In this embodiment, the structure of the transmitting end and the receiving end is the same. When a terminal initiates a connection, it is a transmitting end. When it is connected to the other party under the request of another terminal, it is the receiving end. In this step, a network coding layer is constructed between the transport layer and the application layer of the terminal; the network coding layer is respectively connected to the transport layer and the application layer, and in the connection with the transport layer, the network coding layer uses (or passes) the transmission. The layer's UDP protocol establishes a connection with the transport layer. For the location and data transmission diagram of the above network coding layer, please refer to FIG. 1. In FIG. 1, the solid arrow indicates the data transmission path, and the dotted arrow indicates the transmission path of the response signal (ACK).

Step S12 forms a network code, which is parallelized into a numbered data packet, and is sent by the transmitting end to the receiving end: In this step, one terminal needs to transmit the data or file to another terminal, and then the terminal initiates the connection and sends the data. The other terminal responds to the connection it initiated as the receiving end. The transmitting end transmits the file to be transmitted from the application layer to the network coding layer, and encodes the encoded data to obtain the encoded data; the obtained encoded data sequentially enters the transmission layer of the transmitting end; and processes it in the transmission layer of the transmitting end to obtain With numbered packets, multiple such packets form a block of data that is sent to the network through the transport layer. At the time of the above data transmission, the transport layer transmits data in units of packets, but the network coding layer decodes the encoded data in units of data blocks. In the above transmission process, the setting of the transmission parameters, the transmission of the data packet, and the reception of the acknowledgment signal (sent by the receiving end after receiving the data packet) are included, and the details will be described later.

Step S1 3: The receiving end receives the data packet, obtains its number, and sends a different response signal to the transmitting end according to the numbering condition: After the data reaches the receiving end through the network link, the receiving end receives the data packet, and obtains the data packet in the transport layer. The number of these data packets, according to the relationship between the previously obtained number and the currently obtained number, sends a different acknowledgment signal to the sender, for example, if the obtained packet number is continuous, indicating that there is no packet loss. Appears, sends an acknowledgment signal to the sender; if the number of the packet is not continuous, indicating that there is a packet in the middle, the extension acknowledgment signal is sent to the sender. The extended acknowledgement signal carries the number of the last received continuous data packet and the currently received data packet number, so that the sender can determine which data packets are retransmitted.

Step S14: The transmitting end determines whether to retransmit or not retransmit the related data packet according to different response signals: In this step, the transmitting end determines whether to retransmit or not according to the received different type of response signals sent by the receiving end. Send a data packet, and if it needs to be resent, also determine which data needs to be resent. Package.

In this embodiment, specifically, step S12 further includes multiple steps, as follows:

Step S21: The transmitting end sends the data packet with the transmission parameter to the receiving end, and the receiving end returns an acknowledgment signal and sets it according to the setting: In the network transmission, the receiving end that needs to arrive due to the data is usually indefinite, in order to ensure the transmission The sender needs to notify the receiving end of the transmission parameters before sending the data, that is, when the connection is established, so that the receiving end can set according to these parameters, thereby realizing the connection transmission data. In this step, when the transmitting end establishes a connection with the receiving end, the transmitting end sends a data packet carrying the set transmission parameter to notify the receiving end, and the receiving end returns the acknowledgement information after receiving the data packet and according to the transmission. The parameter sets the receiver.

Step S22: setting or adjusting the data block size according to the specification: In this step, the data or file to be transmitted by the transmitting end is divided into the data packet of the first set length according to the inflow order thereof, when the remaining file or data length is smaller than the first setting When the length is fixed (that is, when an integer number of data packets are obtained according to the first set length, and some data remains undivided, but the length of the data is less than the first set length), the data zero is filled in to keep the data packet length; A fixed number of data packets constitutes a data block, and the data packet is the smallest unit of transmission; the number of data packets included in the data block is adjusted according to the information returned by the receiving end during transmission. That is, the sender divides the stream or file into multiple blocks, including the number of b lks i ze packets, each of which is assumed to be a fixed length. If the remaining files or streams are not enough to form a complete package, pad it with zeros to ensure that all packets have the same length. A block does not need to be full, that is, one block may be less than b lks i ze packets; however, the i-th block must be full before the i+1th block is initialized. After the initial packet transmission, the block size is adjusted according to the feedback from the receiver. The receiver is responsible for decoding the received data and constructing an ACK to the sender. When the receiver receives a packet, it needs to check if the current block is decodable (ack_currb lk) and the number of degrees of freedom (ack-currdof) of the current block received.

For each block b lkno , the receiver side initializes a coding coefficient matrix C _{blkn of} b lks i ze b lks i ze . And a corresponding payload structure P _blkn . . Whenever a packet is received from b lkno, the coding coefficients and payload are inserted into C _blkn respectively. And P _blkn . . The Gaussian elimination method is then used to determine whether the received data is linear regardless of whether the previously received packet is linear. If yes, the receiver sets ack-currdof ack.currdof + 1. If ack_currdof is equal to b lks i ze, the receiver acknowledges that it has received sufficient degrees of freedom for the ack_currb lk block and updates ack_currb lk Ack.currblk + 1 (requires resetting ack_currdof to reflect the need for new ack_currblk degrees of freedom). If the received packet is linearly related to the previously received packet, the receiver transmits an ACK (received by the corresponding packet) but does not update ack_currdof and ack_currblk. Once enough linearly unrelated blksizef packets (degrees of freedom) are received, the receiver can decode all the packets in the block.

Step S23: The sender processes the encoded data according to the set format to obtain the numbered data packet and the data block: In this step, the number of data packets in the data block is adjusted, and in the case that the network environment is not good, the data block is reduced. The number of packets in the packet helps to successfully decode. Therefore, when the network environment is good, the number of data packets in the data block is large; otherwise, the number is small. That is, in this step, the number of data packets included in the data block is adjusted according to the transmission time of each data packet, and the data packet includes the number of data packets and the data packet between the transmitting end and the receiving end. The transmission time is inversely proportional.

Step S24 sets or adjusts the number of tokens according to the regulations: In this step, the number of tokens is set or adjusted, which is also changed according to the network environment, and the sender controls the data packet by the number of tokens allowed to be allowed in a unit time. The sending speed, each token is allowed to send a data packet; the number of tokens appearing per unit time is adjusted according to the time of packet transmission. The number of tokens allowed to be sent in a unit time is adjusted according to a transmission time of the data packet at the transmitting end and the receiving end, and the receiving end is notified; the longer the data packet is transmitted, the transmission is allowed in a unit time. The less the number of tokens.

In the present embodiment, Fig. 3 only gives an example. For the sake of illustration, the steps in Figure 3 are arranged according to certain conditions. However, in actual situations or applications, the data transmitting end may not be arranged in the order shown in FIG. 3. It is possible that the two steps in FIG. 3 are performed simultaneously or the order of the two steps in FIG. 3 is reversed. Or for some reason not doing one of the steps and so on. These situations are all possible in this embodiment. In summary, the steps in Figure 3 are implemented to send data smoothly and reliably, and different steps can be taken depending on the actual situation. For example, in the case of the same pair of transmitting end and receiving end, there must be a parameter setting step at the beginning; however, after a period of transmission, it is necessary to adjust the number of data packets included in the data block, and the above steps are not necessary. All of them are executed once, and it is only necessary to directly adjust the number according to the above description (ie, directly execute step S23).

As described above, in the above step S13, after the data packet is received by the receiving end, it is processed in the transmission layer of the receiving end to obtain the data packet number thereof, and is judged if the number is connected. And sending an acknowledgement response signal to the sending end; if the number is not continuous, sending an extended acknowledgement response signal to the sending end, where the extended acknowledgement signal carries the last consecutive data packet number received by the receiving end .

In addition, the network coding layer of the receiving end also needs to decode the received data packet. It first needs to determine if these packets can be decoded. Refer to FIG. 4 for the judgment process, including the following steps: Step S31: Receive a new data block, initialize the coding matrix and the payload: In this step, when a new data block is received, the initialization size is b Iks i zeXb Iks i ze The coding coefficient matrix and the corresponding payload structure, where blks ize is the length of the currently set data block.

Step S32: The newly received data packet is linearly related to the data packet received before the data block. If yes, go to step S34, otherwise, go to step S35; for each received data packet, insert its coding coefficient and payload respectively. In the above coding coefficient matrix and the corresponding payload structure, the Gaussian elimination method is used to judge whether it is independent of whether the previous data packet is linear.

Step S33: Sending an acknowledgment signal, updating the block degree of freedom: Since the linearity is irrelevant, indicating that the data packet is received normally, the receiving end returns an acknowledgment signal, updates the current block degree of freedom, increments it by 1, and executes step S35.

Step S34 sends an acknowledgment signal, and does not update the block degree of freedom: In this step, the acknowledgment signal is transmitted, but the block degree of freedom is not updated, and the next packet is received, and the process returns to step S32.

Step S35: The degree of freedom is equal to the block length. No: It is judged whether the updated current block degree of freedom is equal to the current set length in the data block, and if so, it is judged that the data block can be decoded; otherwise, the process returns to step S32.

Step S36 The data block is received and can be decoded: In this step, all the data packets in one received data block are decoded. At the same time, it is ready to receive the next data block.

In addition, in this embodiment, if the sending end sends a data packet and the time when the acknowledgment signal is not received is greater than the transmission time of the last data packet plus the retransmission timeout period, the transmitting end transmits the data packet in the slow start mode; The slow start mode includes setting the number of tokens allowed to be transmitted in the unit time to a default value, and adding the number of tokens after successfully transmitting one data packet, until the number of tokens reaches a set threshold, and then changing Normal transmission status.

In summary, in the present embodiment, since the network coding layer operates on the originally unreliable UDP transport layer, it is necessary to consider the case where the message is lost. See Figure 5 for UDP To introduce its efficient and reliable transmission function, its network coding layer protocol needs to consider six states: 3⁄4 port:

CLOSE: This state is in the absence of a connection; LI STEN: Enters the LI STEN state when passively connected, in which it waits for an active connection request; SYN-SENT: When initiated After a SYN that actively establishes a connection, it enters the SYN-SENT state. At this point, a connection status record is generated, the initial sequence number of the connection is determined, and a SYN is sent to the far end, waiting for the opposite SYN and ACK to be acknowledged; SYN-RECVD (SYN receiving): This state can be from the LI STEN state and SYN- The SENT state is converted. When in the LI STEN state, it transitions to the SYN-RECVD state as soon as it receives a far-end SYN segment. And generate an initial sequence number as the starting number of the local message sequence, and send a SYN and ACK to the opposite end, and then wait for the opposite end to respond; OPEN (connection): indicates that the establishment of the connection is successful, both parties negotiate good configuration data as initial Serial number, maximum segment size, maximum send window, etc., then both parties can interact with each other through the virtual connection; CLOSE-WAIT. The CLOSE-WAIT state can be received from the OPEN state c l ose (open) request to switch to CLOSE state or receive RST message. In both cases, enter the CLOSE-WAIT state and wait for a while to close the connection.

When in the SYN-SENT state, when the remote SYN is received but there is no accompanying ACK, it is transferred from the SYN-SNET state to the SYN-RECVD state. This situation is due to the fact that both parties actively initiate a request to establish a connection. At this time, the SYN is sent again with the same serial number, and the SYN of the other party is confirmed, indicating that the connection establishment request is received.

In this embodiment, a virtual connection is also established between the network coding peers of the sender and the receiver, similar to TCP, and a three-way handshake is also performed. According to the actual situation, if both parties initiate a connection at the same time, respond and send a SYN. A SYN packet is sent when the connection is established. The configuration BPDUs of both ends are carried in the segment. In general, closing a connection uses a 4-way handshake similar to TCP, but in a persistent connection application, the connection is closed only when an exception occurs. After an abnormality occurs on one end, the other end detects the abnormality through the reliable detection mechanism and reports it to the upper-layer user to release the local connection resources.

In this embodiment, efficient and reliable transmission is also achieved by means of the following components or mechanisms: Retransmission timer: In the network coding layer, after each coded message is sent, if the retransmission timer is not set at this time , then set the timer. So when the peer receives the message, it will send a response to On the sending end, the sending end cancels the timer after receiving the response. If there is a message sent but not answered after receiving the response, the timer is restarted. The duration of the timer timeout is negotiated by both ends, but both ends must be the same value.

Retransmission counter: The retransmission counter records the number of times the message is retransmitted until the number of retransmissions reaches the threshold. When the number of retransmissions exceeds the threshold, the connection is considered to have been disconnected. In this case, the connection error should be handled.

Response mechanism: There are two types of response messages for encoded messages. EACK and ACK. Consideration must be given to the loss of the intermediate sequence number message after a group of messages has been sent. To this end, this scenario provides an EACK message type. When the receiver receives an unordered message, the receiving end sends an EACK response to the sender, notifying that some of the other party's messages may be lost or corrupted during the flight. When the sender receives the EACK type message, the EACK message carries the most recently received sequence message number and the received out-of-order message number. The message sequence number that needs to be retransmitted is calculated according to the information carried in the message. Resend these messages. Similar to TCP to reduce network load and increase utilization. It also uses response delay techniques. The receiver maintains a counter to count the number of received but unacknowledged packets. When the received packet exceeds the threshold (the threshold is configurable), a separate ACK is sent. An EACK is sent if the message does not respond. In addition, there is a cumulative timeout timer to control the waiting time. When the cumulative acknowledgment timer expires, a separate ACK or EACK is sent regardless of whether the accumulated counter has reached the threshold. The timer restarts when a separate response is sent.

Keep-alive mechanism: When transmitting information using the network coding layer protocol, there may be a case where there is no data circulation on one connection for a period of time. That is to say, neither link has applied data to the other party. At this point, a keep-alive mechanism is needed to keep abreast of the link situation so that it can be detected in time if one party crashes. When one end of the connection finds that there is no message transmission on the local end, the keepalive timer is set on the local end. If there is data transmission before the keep-alive timer expires, the keep-alive timer is canceled and the re-send timer is set. If the keep-alive timer expires and no application data needs to be sent, a keep-alive message is sent to the peer. The keep-alive mechanism enables the network coding layer to provide long connection services.

Redundant connection mechanism: If a connection fails, the upper application receives the signal and starts the state transition timer. The upper application can transfer the connection status information through the API, initiate another connection and inherit the previously failed connection. This avoids message loss or duplication. If the state transition timer has not expired yet, the disconnected connection resource is released. In this embodiment, a message of type SYN is used to initiate a connection between two hosts. The SYN message also contains the configuration negotiation information required by both parties. For example, various timer durations, version numbers, connection IDs, message sequence spaces, cumulative acknowledgement thresholds, and so on. Messages of type SYN cannot carry user data. The ACK is used to acknowledge the received in-sequence message to the sender. EACK, that is, extended ACK, when the sender receives the EACK, it indicates that a message is lost in the transmission, and the receiving end receives the out-of-order message. The EACK always contains a sequence message sequence number and one or more out-of-order message sequence numbers that were last received last. After receiving the EACK, the sender will retransmit the last sequence message to the message before the out-of-order message (excluding the two sequence numbers) according to the contents of the EACK. As described above, in the embodiment,

In this embodiment, the network coding layer uses tokens to control the transmission rate of the sender instead of the congestion window cwnd; therefore, tokens play an important role in the network coding layer as if cwnd is TCP-like. A token allows the sender of the network coding layer to send a packet (coded or unencoded). When the sender transmits a packet, tokens can be used. The number of Tokens is controlled based on the modified AIMD multiplication compensation:

o _ RTT匪

RTT

RTT = RTT _max when the packet loss occurs (link queue is full). In this case, when a link is provided with a buffer with a bandwidth delay product, then RTT _max = 2 RTT _min , β = 0.5, which is the same as standard TCP. In general, when a queue overflow occurs, the sum of the stream throughputs must equal the capacity of the link toke _{= B} ^ where _n is the number of streams. After compensation according to Algorithm 3, the queue is cleared, throughput = _{1 The} sum of the _RTT quantities becomes ° = i ^Pi _R ^to _T ^k _T ^mS = B . That is, in the above selection, β reduces the number of tokens, so that the link queue is emptied and the throughput is maintained. In the lossy link (except for the loss of queue overflow), RTT is used to adapt β to each loss. When a network path utilization is low, RTT = RTT _min (hence, β = 1 and β * token=tokens). Therefore, tokens do not reduce packet loss, although tokens can be increased in the presence of packet loss. Once The link begins to experience a queuing delay, then RTT> RTTmin, β < 1, ie the tokens are reduced when there is a packet loss, because the link queue is full, and the sum of the throughput before the loss is ∑ ⁿ = B. After the tokens are reduced, when the queue is emptied, the sum of the throughputs is at least 5; ^^ = β (all stream compensation tokens), that is, β can be adjusted to maintain full throughput.

In this embodiment, the parameters of the sender are defined as follows:

In this embodiment, the update of the network parameters of the sender is that the sender adjusts its coding and congestion control according to some sender parameters. If the sender does not receive the sender's transmission within a certain period of time

ACK, the parameter is reset to the default value. After receiving the ACK, the parameters are updated according to the ACK. t ime — las tack Is the current time current time; If the smallest undecoded block (number) is greater than the current sender's block number, you need to send the smallest undecoded block to the receiver, so the current send block number is updated to ack_currblk; in fact, currdof is always ack_currdof and The larger the currdof, the greater the freedom to perform various operations on the packet. For example, the network coding can operate the data bits in the data packet, such as "bitwise XOR", etc., which can effectively allow the target node to accept multiple information without increasing the number of data packets it can receive, that is, without increasing the overall network. Capacity, and bitwise to get more freedom.

For sequence number 1, when the confirmed ack.seqno is greater than seqno.una (unconfirmed), it indicates that there is a packet loss (because the confirmation is in order, from small to large, the small is always confirmed earlier than the large number) The average loss rate is shown in Algorithm 1, and its seqno.una is updated to ack.seqno.

The method for the sender to control its network congestion can be slow start (start slow-start) and congestion avoidance; slow start initially set the congestion window cwnd=l, so that the sender only sends one segment at the beginning, and then gradually increases Cwnd, this is much more than a large cwnd - the next sub-injection of many segments into the network.

Initially, if current time > time_lastack + RTO , the time when the ACK is received on the time axis plus the retransmission timeout period is less than the current time, it means that the retransmission acknowledgement message cannot be received even if the packet is lost or the like. So start the slow start mode.

When an ACK is received, at the slow start, the token is incremented by 1, indicating that the network coding layer sender increases the number of packets allowed to be sent at one time. When the token tokes is increased to the slow start threshold ss_threshold, the slow start is used and the congestion avoidance algorithm is used instead. . The number of tokens is controlled according to the modified AIMD. That is, use the above method and apply tokens to implement congestion control.

For the receiving end, it is mainly based on the degree of freedom in ack_currblk whether the controller can be decoded. If the received block is linearly correlated, it means that the received block is not used as the decoded coefficient matrix. Once there are enough linearly independent blocks, each block can be represented by these linearly independent blocks, so it can be decoded.

For each block (blkno), the receiver side initializes a coding coefficient matrix C _{blkn of} blks ize _blksize . And a corresponding payload structure P _blkn . . Whenever a packet is received from blkno, the encoding coefficients and payload are inserted into C _blkn respectively. And P _blkn . . The Gaussian elimination method is then used to determine whether the received data is linear regardless of whether the previously received packet is linear. If yes, the receiver sets Ack-currdof ack.currdof + 1. If ack_currdof is equal to blksize, then the receiver acknowledges that it has received enough ack_currblk blocks and updates ack_currblk ack.currblk + 1 (requires resetting ack.currdof to reflect the new ack.currblk degrees of freedom). If the received packet is linearly related to the previously received packet, the receiver transmits an ACK (received by the corresponding packet) but does not update ack_currdof and ack_currblk. Once enough linearly unrelated blksizef packets (degrees of freedom) are received, the receiver can decode all the packets in the block.

If the block blkno is C _blkn . [ index, :] is empty, that is, the data packets in the block are linear and independent, no iteration is required, and TRUE is returned. The coding coefficient and the load normalized by the block blkno constitute C _blkn . And P _blkn .

If the index entry is smaller than the size of the block, indicating that there is a linearly related packet in the block, it returns FALSE. It is necessary to iteratively update the parameters so that the linear correlation coding coefficient is 0. Once updated, it needs to be normalized once.

However, it is not to be construed as limiting the scope of the invention. It should be noted that a number of variations and modifications may be made by those skilled in the art without departing from the spirit and scope of the invention. Therefore, the scope of the invention should be determined by the appended claims.

Claims

claims

1. A network coding and transmission method based on UDP protocol, which is characterized by including the following steps:

A) Construct a network coding layer between the transport layer and the application layer; the network coding layer is connected to the transport layer and the application layer respectively, and the network coding uses the UPD protocol of the transport layer to communicate with the transport layer establish connection;

B) Data or files enter the network coding layer of the sending end from the application layer of the sending end, cache and perform network coding on the sending end of the network coding layer, form them into numbered data packets according to the set method, and send them through the transmission layer of the sending end. to the receiving end;

C) The network coding layer of the receiving end receives the encoded data through the transmission layer of the receiving end, identifies the number of the received data packet, returns different response signals to the sending end according to the received number, and simultaneously decodes it to obtain the data. or file, and transmitted to the application layer of the receiving end.

2. The method for network coding transmission based on UDP protocol according to claim 1, characterized in that it further includes the following steps:

D) The sending end decides to resend or not resend the data block based on the different types of response signals received from the receiving end.

3. The network coding and transmission method based on UDP protocol according to claim 2, characterized in that step B) further includes the following steps:

B1) When the sending end and the receiving end establish a connection, send a data packet carrying set transmission parameters to notify the receiving end. After receiving the data packet, the receiving end returns confirmation information and follows the transmission parameters. Set this receiver.

4. The network coding and transmission method based on UDP protocol according to claim 3, characterized in that step B) further includes the following steps:

B2) The sending end divides the data or files to be transmitted into data packets of the first set length in the order in which they flow in. When the length of the remaining files or data is less than the first set length, data zeros are filled in to maintain The length of the data packet makes each divided data packet have the same length; the number of data packets set to blksize constitutes a data block, and the data packet is the smallest unit of transmission; the number of data packets contained in the data block is transmitted The process is adjusted based on the information returned by the receiving end.

5. The network coding and transmission method based on UDP protocol according to claim 4, wherein Characteristically, step B) further includes the following steps:

B3) The sending end controls the sending speed of the data packet through the number of tokens allowed to appear per unit time, and each token is allowed to send one data packet; the number of tokens appearing per unit time is based on the data packet transmission time adjustment.

6. The network coding and transmission method based on UDP protocol according to claim 5, characterized in that the number of tokens allowed to be sent per unit time is based on the transmission time of one data packet between the sending end and the receiving end. Adjust and notify the receiving end; the longer the data packet transmission time is, the smaller the number of tokens allowed to be sent per unit time.

7. The network coding and transmission method based on UDP protocol according to claim 6, characterized in that, the number of data packets included in the data block is adjusted according to the transmission time of each data packet, and the number of data packets included in the data block is adjusted The number is inversely proportional to the transmission time of the data packet between the sender and the receiver.

8. The network coding and transmission method based on UDP protocol according to claim 7, characterized in that step C) further includes the following steps:

C1) The receiving end detects the received data packet number, and if the numbers are continuous, sends a confirmation response signal to the sending end; if the numbers are discontinuous, sends an extended confirmation response signal to the sending end, the extended confirmation signal carries the last consecutive data packet number received by the receiving end.

9. The network coding and transmission method based on UDP protocol according to claim 8, characterized in that step C) further includes the following steps:

C2) When receiving a new data block, initialize the coding coefficient matrix and the corresponding payload structure with a size of blksize*blksize, where blksize is the length of the currently set data block;

C3) For each received data packet, insert its coding coefficient and payload into the above-mentioned coding coefficient matrix and corresponding payload structure respectively, and use the Gaussian elimination method to determine whether it is linearly independent of the previous data packet. If so, Go to step C4); Otherwise, go to step C5);

C4) Returns a confirmation signal, updates the current block degree of freedom, increments it by 1, and determines whether the updated current block degree of freedom is equal to the current set length in the data block. If so, determines that the data block can be decoded; otherwise, wait for reception next packet;

C5) sends an acknowledgment signal but does not update the block degrees of freedom and waits to receive the next data packet.

10. The network coding and transmission method based on UDP protocol according to claim 9, characterized in that if the sending end sends a data packet and does not receive the confirmation signal for a time greater than the transmission time of the last data packet plus retransmission timeout period, the sending end uses the slow start mode to transmit the data packet; the slow start mode includes setting the number of tokens allowed to be transmitted per unit time to a default value, and the number of tokens is increased by each time a data packet is successfully transmitted. 1. Until the number of tokens reaches the set threshold, it changes to the normal transmission state.