JP2006222908A - Retransmission method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a new retransmission method that is an alternative to TCP/IP as a retransmission method that is suitable for stream data distribution. <P>SOLUTION: Receivers 206, 207 check for error in each of 64 KB packet received from transmitters 202, 203 via a packet communications network 201, and return existence or nonexistence of bit error or packet loss in each packet to the transmitters 202, 203 with an ACK packet. If the transmitters 202, 203 detect the occurrence of the error, on the basis of the received ACK packet at this point, it reads out the packet from a retransmission buffer. During the reading time of the retransmission packet reading, writing to a next video signal packet buffer is performed. Consequently, transmitting of the retransmission packet is completed before the next video signal is packetized and transmitted. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a retransmission method used for stream transmission for transmitting stream data via a packet communication network.

  With the widespread use of large-screen liquid crystal televisions, plasma televisions, and the like, there is a demand for higher-definition video quality for television broadcasting (referred to as TV broadcasting). In addition, high-definition broadcasts have already been broadcast, but with the start of terrestrial digital TV broadcasts, TV broadcasts are scheduled to be sequentially transferred to digital high-definition broadcasts (HDTV).

  Since the HDTV is a digital system, it is not easily affected by the weather, and the image quality (resolution) is remarkably improved as compared with conventional analog TV broadcasts.

  Such a trend toward higher image quality affects not only the field of TV broadcasting, but also video distribution on the Internet. For example, in the field of movies, that is, digital cinema as one of high-quality contents in video media, a display method with higher quality than the resolution of HDTV is required. In response to such a request, transmission experiments for digital cinema distribution have been conducted in various directions. As an example, the “Super Multi-connection using TCP multi-connection” described in IEICE, IEICE Tech. A long distance and high speed stream transmission experiment of high definition moving images will be described.

  This system is a SHD (Super High Definition) digital cinema distribution system with a resolution of 3840 x 2048 (approximately 8 million) pixels, RGB of 10 bits, and 24 fps display per second. It has four times the resolution of HDTV and exceeds 35 mm film. Image quality screening is possible. 35mm original negative or interpositive film is scanned frame by frame and used as moving image data. Since the bit rate of an SHD moving image is about 5.7 Gbps without being compressed, the SHD moving image is compressed to about 50 Mbps to 300 Mbps without causing visual image deterioration, and is stored in the cinema distribution server together with uncompressed PCM audio data. JPEG2000 is used for image compression. The moving image and audio data are TCP stream-transmitted through an IP network using Gigabit Ethernet (registered trademark) in response to a request from a cinema client (decoder), and output from a projector and a speaker.

  As a result of conducting a transmission experiment on an IP network covering 3000 km using this system, stream transmission of only 50 Mbps data was completed. Further, for parallel stream transmission with 64 connections, 200 Mbps stream transmission was possible.

  RTT (Round Trip Times) in this system was 59 msec, the TCP window size was expanded to 4 MB, and the theoretical value of throughput was set to be 570 Mbps. As a result, only a throughput of 50 Mbps was obtained. According to the above document, the difference from this theoretical value is because the congestion window size is not sufficiently widened on the transmission side by the original congestion avoidance algorithm of the Linux TCP protocol stack.

In addition, in data transmission using TCP multi-connection, since the restriction by the unique congestion avoidance algorithm is performed for each TCP socket, the substantial congestion window size increases, which is considered to greatly contribute to the increase in throughput. ing. Further, it is described that by suppressing the burst property of transmission data, it is possible to suppress buffer congestion in addition to reducing the load on the network, thereby improving throughput.
The Institute of Electronics, Information and Communication Engineers, IEICE Tech.

  However, as described in the above-mentioned document, various problems have been pointed out in the case of the stream transmission method using TCP / IP, such as the inability to obtain the throughput as the theoretical value. As a first problem, regarding the transmission traffic pattern of the stream data controlled by the operating system (OS), since the peak rate is very high and the transmission rate is close to the transmission rate, the transmission is performed on the network. There is a problem that packet loss occurs. As a next problem, it is pointed out that the congestion window size is not sufficiently widened on the transmission side by the unique congestion avoidance algorithm of the TCP protocol stack. In addition, although not described in the above-mentioned document, on the actual network, the RTT varies greatly depending on the traffic condition or the distance of the communication partner, so that it is conceivable that the arrival of packets is delayed and the throughput is lowered. In addition, a decrease in transmission rate when retransmission processing is performed due to packet loss due to a transmission error on the transmission path is also a serious problem.

  In the transmission experiment described in the above document, proposals for improving the throughput are made by various efforts. However, in stream transmission using a general-purpose OS or TCP / IP, the speed of the transmission path On the other hand, it is expected that the transmission rate of the transmission data is limited to be considerably low.

  In the future, when a high-quality video data distribution system is constructed, a new real-time retransmission method other than the TCP / IP protocol is required.

  Therefore, an object of the present invention is to provide a new retransmission method that replaces TCP / IP as a retransmission method suitable for distribution of stream data.

  In order to achieve the above object, the present invention is a retransmission method used for transmission of stream data through a packet communication network, and performs communication while ensuring a transmission band of a transmission packet and a retransmission packet within a certain period, A retransmission method is provided, wherein when the error occurs in the packet communication network, the retransmission packet is retransmitted within the fixed period.

  According to the present invention, a new retransmission scheme can be provided as a retransmission scheme suitable for the distribution of stream data. For example, when high-quality video data is distributed in real time, even if a transmission error occurs on the network, high-quality video can be displayed on the monitor without causing disturbance in the video data. More specifically, stream transmission up to the transmission speed of the network becomes possible.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  FIG. 1 is a block diagram showing a network configuration to which a retransmission method according to an embodiment of the present invention is applied.

  In the network, as shown in FIG. 1, a plurality of transmission devices 202 and 203, a PC (personal computer) 205, a plurality of reception devices 206 and 207, a management device 210, and the like are directly connected to the packet communication network 201. The packet communication network 201 is composed of, for example, Ethernet (registered trademark) with a transmission rate of 10 Gbps. Although not shown, the packet communication network 201 is directly connected to a number of PCs and network devices other than those described above.

  The transmission devices 202 and 203 receive a DVI (Digital Visual Interface) format video signal from a video source, convert the video signal into a 10 Gbps packet, and transmit the packet to the packet communication network 201. Here, the transmission apparatus 202 is connected to a PC 204 that forms a video source. The PC 204 is equipped with a graphic card provided with a DVI terminal for outputting a video signal and a serial I / F, and the DVI terminal and the serial I / F are each connected to the transmission device 202. The transmission device 203 is connected to a PC 205 serving as a video source, and the PC 205 is equipped with a graphic card provided with a DVI terminal for outputting a video signal. The PC 205 is connected to the packet communication network 201 as described above.

  Receiving devices 206 and 207 receive packet signals via the packet communication network 201, convert the received packets into DVI format video signals, and output the signals to the monitors 208 and 209. The monitors 208 and 209 have, for example, a resolution of 1600 × 1200 (UXGA).

  The management device 210 monitors and manages the traffic of the packet communication network 201, and issues an instruction for designating a communication path and the like when performing video communication.

  Next, a video signal retransmission method in the network will be described with reference to FIG. FIG. 2 is a diagram showing video signal transmission / reception timing in the network of FIG.

  In each of the transmission devices 202 and 203, the DVI-format video signal from the PCs 204 and 205 is converted into a 10 Gbps packet. Here, for convenience of explanation, it is assumed that the data rate of the video signal is 4 Gbps. The transmission devices 202 and 203 include a packet buffer and a retransmission buffer having a size of 64 KB (window size), for example.

  The transmitters 202 and 203 write the video signal from the PCs 204 and 205 into the packet buffer for 64 KB (see FIG. 2A), and then send the video signal from the packet buffer to 10 Gbps which is the transmission speed of the packet communication network 201. Read at speed. The read video signal is divided into 1.5 KB lengths, converted into a packet format (Ethernet (registered trademark) packet format) of the packet communication network 201, and sent to the packet communication network 201 (FIG. 2B). )). That is, a 4-Gbps video signal is converted into a 10-Gbps packet every 64 KB period (= 128 μs). At this time, the data length is converted from 128 μs to 51.2 μs, and video packets are continuously transmitted with a packet interval of 76.8 μs.

  Here, it is assumed that a packet transmitted to the packet communication network 201 has a bit error or packet loss in the packet communication network 201. The receiving devices 206 and 207 perform error checking for each received 64 KB packet, and return the presence or absence of a bit error or packet loss for each packet to the transmitting devices 202 and 203 using an ACK packet (FIG. 2 (c)). See). Here, the packet is transmitted from the transmission devices 202 and 203 to the reception devices 206 and 207 through the packet communication network 201, until a packet error is detected, an ACK packet is sent back to the transmission devices 202 and 203, and an ACK packet is detected. The maximum time RTT (Round Trip Times) is known in advance by measurement or the like. Here, RTT = 12.8 μs, and the transmitting apparatuses 202 and 203 can detect whether or not an error has occurred in the transmitted packet within that time.

  The transmission devices 202 and 203 store the transmission packet in the retransmission buffer. When it is detected that no error has occurred in the packet transmitted based on the received ACK packet, the retransmission device clears the retransmission buffer and an error occurs. When it is detected that the packet is being transmitted, the packet is read from the retransmission buffer. This retransmission packet is read out over a time period of 51.2 μs. During the reading period, the next video signal is written to the packet buffer, so the next video signal is packetized. Transmission of the retransmission packet ends. The time from the retransmission packet until the next packet is transmitted has a margin of 12.8 μs, and during this time, data other than the video signal can be transmitted. The receiving devices 206 and 207 temporarily store the received video packets in the reception buffer. If there is an error in the packet, the receiving devices 206 and 207 discard it, and if there is no error, it is the transmission rate to the monitors 208 and 209. A packet is read from the reception buffer at a rate of 4 Gbps (see FIG. 2D). In this way, video signals can be output to the monitors 208 and 209. If there is an error in the retransmission packet, the packet is discarded or the error packet is converted into a video signal as it is without making a retransmission request.

  Next, a case where the data rate is 2.5 Gbps will be described with reference to FIG. FIG. 3 is a diagram showing video signal transmission / reception timings when the data rate is 2.5 Gbps in the network of FIG.

  When the data rate is 2.5 Gbps, the operation is basically the same as described above. However, when there is an error or packet loss in the retransmission packet, the retransmission process can be further performed. As shown in FIG. 3, since the data length of a 64 KB video signal is as large as 204.8 μs, a band capable of transmitting three packets having a packet length of 51.2 μs is generated, and two retransmissions are performed. (See FIGS. 3 (a) and (b)). That is, when an error occurs in the retransmission packet and the reception devices 206 and 207 detect it, an ACK packet is returned, and a second packet retransmission is performed in response to the reception of the ACK packet. However, if an error or packet loss occurs in the second retransmission packet, the packet is discarded or the error packet is converted into a video signal as it is without making a retransmission request (FIGS. 3C and 3D). See).

  In this way, even if an error occurs in the video signal, the video can be reproduced without causing disturbance in the video by retransmitting the packet within a predetermined time.

  Next, settings regarding video transmission will be described.

  First, determination of the window size and the number of retransmissions will be described.

  Prior to the communication, communication arbitration is performed between the PCs 204 and 205 which are video sources and the transmission devices 202 and 203. The PCs 204 and 205 start the video transmission application and notify the transmission apparatuses 202 and 203 of the communication partner and the communication rate. Here, when the communication partner is a monitor capable of displaying UXGA as in the present embodiment, for example, a transmission rate of 4 Gbps corresponding to the resolution is notified to the transmission apparatuses 202 and 203 as a communication rate. For example, in the case of the PC 204, the PC 204 directly sends information including a communication partner and a communication rate to the transmission device 202 via the serial I / F. In the case of the PC 205, the above information is once sent from the PC 205 to the management device 210 that manages network traffic via the packet communication network 201. Then, the management device 210 notifies the transmission device 203 of the above information. Thus, the notification of information including the communication partner and the communication rate from the PCs 204 and 205 to the transmission devices 202 and 203 may be performed in any of the above forms, and if the information can be notified, The form is not asked.

  The transmission apparatuses 202 and 203 determine the window size Ws (packet buffer storage unit size) and the number of retransmissions N based on the notified communication rate using the following equation (1).

Ws / Vp> (N + 1) × Ws / Vn + N × RTT (1)
Ws: Window size → 64KB
N: Maximum number of retransmissions Vp: Data rate → 4 Gbps
Vn: Network transmission rate → 10 Gbps
RTT: Round Trip Times → 12.8μs
At this time, the RTT is a value for each communication partner notified in advance from the management device 210 or the like, and the value is retrieved from the communication partner. Alternatively, the RTT information may be included in the communication information and notified from the PCs 204 and 205 or the management apparatus 210 each time.

  Here, the relationship between the window size and the number of retransmissions will be described. In the transmission apparatuses 202 and 203, as described above, the video signal is processed in units of 64 KB packets. That is, in the case of hardware with a fixed window size, the number of retransmissions is varied according to the data rate. From the above equation (1), it can be seen that the number of retransmissions is 1 when the data rate is 4 Gbps, and the number of retransmissions is 2 when the data rate is 2.5 Gbps. On the other hand, in the case of hardware that can vary the window size, the number of retransmissions can be fixed. When the packet communication network is a wired network and the network scale is relatively small, the transmission quality is good. For example, when the error rate is about 1.0E-15, there is a practical problem if the number of retransmissions is one. Absent. Further, if the number of retransmissions is fixed, there is an advantage that the window size and the RTT can be increased as the data rate is smaller, and the network design becomes easier. As described above, the window size and the number of retransmissions are determined.

  Next, securing of the communication band by the transmission devices 202 and 203 will be described.

  The transmission devices 202 and 203 secure a bandwidth for the packet communication network 201. When the packet communication network 201 is an Ethernet (registered trademark) network, the packet communication network 201 is configured by an Ethernet (registered trademark) switch, but all the Ethernet (registered trademark) switches through which packets pass to the communication partner are included. On the other hand, bandwidth is secured. In this method, for example, a VLAN (Virtual Local Area Network) is used so that the same VLAN is set to the port of the communication path, and other communication traffic is set not to flow into the port. The switch setting may be performed by instructing the relevant switch from the management apparatus 210 to update the VLAN table. In this way, the communication band is secured.

  Regardless of the method described above, other methods may be used as long as transmission bandwidths for transmission packets and retransmission packets are secured. Also, it is necessary to make RTT constant or guarantee within a predetermined period. For example, if a cut-through type Ethernet (registered trademark) switch is used, the delay becomes constant regardless of the packet size. Can be suppressed. Further, if an ACK reply is surely performed within a predetermined period in the receiving apparatus, the RTT is guaranteed at a constant or within a predetermined period, and the retransmission method of this method can be used.

  Further, when the transmitting device and the receiving device are directly connected, the entire band is already secured, so that the procedure described above need not be taken.

  As described above, according to the present embodiment, when delivering a high-quality video signal in real time, even if a transmission error occurs on the network, the high-quality video is displayed on the monitor without causing any disturbance in the video signal. can do. More specifically, stream transmission up to the transmission rate of the network becomes possible. Due to such characteristics, a new retransmission method replacing TCP / IP can be provided as a retransmission method suitable for distribution of stream data.

It is a block diagram which shows the structure of the network to which the data retransmission system which concerns on one embodiment of this invention is applied. It is a figure which shows the transmission / reception timing of the video signal in the network of FIG. It is a figure which shows the transmission / reception timing of a video signal when a data rate is 2.5 Gbps in the network of FIG.

Explanation of symbols

201 packet communication network 202, 203 transmission device 204, 205 PC (personal computer)
206, 207 Receiving device 208, 209 Monitor 210 Management device

Claims (8)

A retransmission method used for transmission of stream data via a packet communication network,
Retransmission method characterized in that communication is performed while securing a transmission band of a transmission packet and a retransmission packet within a fixed period, and the retransmission packet is retransmitted within the fixed period when an error occurs in the packet communication network .   The retransmission method according to claim 1, wherein the number of retransmissions of the retransmission packet is determined according to a communication rate. 3. The retransmission method according to claim 2, wherein a window size which is a retransmission buffer size is fixed, and the number of retransmissions is set to a number satisfying the following expression.
Ws / Vp> (N + 1) × Ws / Vn + N × RTT
Ws: Window size (retransmission buffer size) (= fixed value)
N: Maximum number of retransmissions Vp: Data rate Vn: Network transmission rate (= fixed value)
RTT: Round Trip Times (= fixed value)   2. The retransmission method according to claim 1, wherein a window size which is a retransmission buffer size is determined according to a communication rate. 5. The retransmission method according to claim 4, wherein the number of retransmissions is fixed, and the window size is set to a size satisfying the following expression.
Ws / Vp> (N + 1) × Ws / Vn + N × RTT
Ws: Window size (retransmission buffer size)
N: Maximum number of retransmissions (= fixed value)
Vp: data rate Vn: network transmission rate (= fixed value)
RTT: Round Trip Times (= fixed value)   2. The retransmission method according to claim 1, wherein a communication path on the packet communication network is set.   7. The retransmission method according to claim 6, wherein the communication path of the packet communication network is set using a VLAN (Virtual Local Area Network) function.   The response signal is returned within a predetermined time using a cut-through type Ethernet (registered trademark) switch whose delay does not change depending on the packet length as means for guaranteeing the RTT constant or within a predetermined period. 3 or the re-transmission method described.

Images