JP7759766B2 - Packet forwarding device and program - Google Patents

Description

The present invention relates to a packet forwarding device and a program.

Until now, program production systems have been constructed using SDI (Serial Digital Interface). In recent years, program production systems have been constructed to replace SDI with standard Ethernet®/IP (Internet Protocol) networks. In an Ethernet®/IP network, devices are connected using a network switch, such as the one shown in Figure 8, as a packet forwarding device. Compared to SDI, an IP network can transmit a larger amount of data per unit time, and different types of signals, such as video and audio, can be multiplexed and transmitted over the same network. As a result, it is suitable not only for the traditional production of HD (High Definition Video) programs, but also for the production of higher resolution programs such as 4K and 8K.

SMPTE ST 2110-20 is a standard for converting video into IP packets and transmitting them over IP networks (Non-Patent Document 1). In addition, SMPTE ST 2110-30/31 has been published as a standard for audio transmission. In these SMPTE ST 2110-20/30/31, video signals are transmitted using UDP (User Datagram Protocol)/RTP (Real-Time Transport Protocol).

P.J. Brightwell, J.D. Rosser, R.N.J. Wadge, P.N. Tudor, "BBC Research & Development White Paper 268," The IP Studio, 2013.

A network switch determines the output port to output a packet from based on the information in the header of the received packet and the state of the network switch. The lines that transmit packets have a limit to the amount of packets that can be transmitted per unit time. Therefore, operators typically estimate the amount of video and audio packets to be transmitted in advance and carefully control transmission equipment such as network switches to prevent the output packets from exceeding the bandwidth supported by the output port (a condition known as congestion). However, congestion can occur, for example, when (1) packets received by multiple ports are output from the same output port, (2) an input port with a faster bandwidth than the output port outputs received packets, or (3) both (1) and (2) occur. Congestion causes packets to be discarded within the network switch.

Figure 8 shows an example of situation (1). Receiving device A is receiving video V1 output from output port #1 of the network switch. During this reception, congestion occurs when another receiving device B starts receiving another video V2 using the same output port #1. The network switch discards some of the packets for video V1 and video V2 to fit the packet output within the bandwidth of output port #1.

When video is transmitted using a transport protocol such as SMPTE ST2022-6 or SMPTE ST2110-20 that uses UDP/RTP and does not use FEC (Forward Error Coding), discarding packets means that the receiving device is unable to obtain all of the necessary pixel information. This results in a degradation of the quality of the video played on the receiving device. For example, the receiving device may experience a frozen output, continuing to output the last successfully received video frame, outputting an image that is clearly an error (for example, all black), or no output at all.

Once congestion like the one described above occurs, it will continue unless the sender stops video output or the network switch stops specific video transfers. Because discarding video and audio packets due to congestion results in degradation of video quality, efforts have been made to prevent video and audio packets from being discarded. For example, network switches with specifications used for video transmission have the ability to sort packets based on the source IP address, destination IP address, source L4 port, and destination L4 port in the IP header, as well as the value of the DSCP (Differentiated Services Code Point) field in the IP header, and assign them to a desired transmission queue. This function pre-configures the network switch with a filter using the address information of the video and audio flows to be transmitted, and assigns packets that match the filter to a specific transmission queue. This prevents packets to be protected from being discarded, while controlling the output data rate of the transmission queue to be equal to or greater than the data rate of the pre-configured video and audio flows.

In addition, multicast is often used for video and audio transmission in program production. Therefore, one method is to discard requests if the payload of an IGMP (Internet Group Management Protocol) packet requesting multicast transmission does not contain a pre-registered multicast address.

However, because program production involves transmitting many video and audio flows, it is difficult to register the IP addresses and port numbers of each in advance. Furthermore, if there is an error in the settings, problems may occur, such as packets being discarded or requests not being accepted and no transmission taking place at all.

SMPTE ST2022-7 includes a method for transmitting the same video packets over two systems, system A and system B, to deal with momentary packet loss. The receiving device receives packets from both system A and system B and performs hitless selection. However, systems often have similar systems for system A and system B. In such cases, congestion occurs at the same time on both systems, resulting in continuous congestion. Because the receiving device cannot receive all packets, it is difficult to assemble the video frame.

The present invention was developed in consideration of these circumstances, and aims to provide a packet forwarding device and program that can reduce the impact on video playback even when congestion occurs during video packet forwarding.

[1] One aspect of the present invention is a packet forwarding device comprising: a receiving unit that receives packets containing data constituting each frame of video; an output unit that outputs the packets received by the receiving unit; an output control unit that determines the priority of the packets on a frame-by-frame basis when the bandwidth of the packets output from the output unit exceeds a threshold; and a priority control unit that discards at least some of the packets received by the receiving unit that contain data of the frame to which a low priority has been assigned by the output control unit, and outputs the packets that have not been discarded to the output unit.

[2] One aspect of the present invention is the above-mentioned packet forwarding device, further comprising a determination unit that determines a flow based on address information set in the packets, and a measurement unit that measures the input speed of the packets to the receiving unit, and the output control unit calculates a thinning rate, which is the rate at which frames are thinned out, based on the input speed measured by the measurement unit and the threshold, and assigns a low priority to some of the frames for each flow according to the thinning rate so that periods during which low priority is assigned are distributed among the flows.

[3] One aspect of the present invention is the packet forwarding device described above, wherein the output unit has a plurality of output interfaces that output the packets to the line, the measurement unit measures the input speed of the packets to the receiving unit for each flow, and the output control unit performs the following processes for each output interface: calculating a thinning rate based on the sum of the input speeds of the flows that output the packets from the output interface and a threshold value corresponding to the output interface; and assigning a low priority to some of the frames for each flow according to the thinning rate so that periods during which a low priority is assigned are distributed among the flows that output the packets from the output interface.

[4] One aspect of the present invention is a program for causing a computer to execute an output control step of determining the priority of packets on a frame-by-frame basis when the bandwidth of packets received by a receiving unit, the packets containing data constituting each frame of video, and output from an output unit exceeds a threshold; and a priority control step of discarding at least some of the packets received by the receiving unit that contain data of the frame to which a low priority was assigned in the output control step, and controlling the packets that were not discarded to be output to the output unit.

This invention makes it possible to reduce the impact on video playback even when congestion occurs during video packet transfer.

1 is a configuration diagram of a network switch according to an embodiment of the present invention. FIG. 10 is a diagram showing flow state information according to the same embodiment. FIG. 10 is a diagram showing a thinning table according to the same embodiment. FIG. 10 is a flowchart showing a transfer control process of the network switch according to the embodiment. FIG. 10 is a flowchart showing a transfer rate measurement process of the network switch according to the embodiment. FIG. 10 is a flowchart showing a QoS value determination process of the network switch according to the embodiment. FIG. 10 is a flowchart showing an entry deletion process of the network switch according to the embodiment. FIG. 1 illustrates a conventional network switch.

Embodiments of the present invention will now be described in detail with reference to the drawings. The packet forwarding device of this embodiment can be applied to, for example, an IP program production system that converts video and other material data used in program production into IP packets and produces programs while exchanging the material data in real time over an IP network.

The packet forwarding device of this embodiment performs QoS (Quality of Service) control that takes video frames into account when the transferable bandwidth is exceeded due to the addition of other traffic while video traffic is being transferred. This control reduces video corruption that can occur when a device receiving video traffic distributed from the packet forwarding device plays the video.

In SMPTE ST2022-6, ST2110-20, and other standards, the number of packets transferred per unit time for each IP flow is determined by the video format. In other words, unless the video format (such as resolution or frame rate) or the packet division method is changed, the number of IP packets containing data that make up one frame remains the same. Conventional packet forwarding devices often randomly discard packets when congestion occurs. If any of the packets that make up a frame are discarded, that frame cannot be reconstructed. However, even if congestion occurs, if all packets that make up a frame can be collected, that frame can be reconstructed. Therefore, the packet forwarding device of this embodiment utilizes this to output packets in a way that reduces degradation of video quality even when the transfer rate is exceeded. Specifically, if the transfer rate of packets output from an output interface exceeds the transfer capacity of that output interface, the packet forwarding device determines whether packets can be forwarded on a video frame-by-video frame basis and sets a QoS value for each packet depending on whether the video frame can be forwarded. This allows video frames to be constructed, albeit intermittently, at the receiving device. This prevents long-term freezes, blackouts, and video interruptions. At the very least, viewers will know what video is being output.

Figure 1 is a block diagram showing the configuration of a network switch 1 according to an embodiment of the present invention. The network switch 1 is an example of a packet forwarding device. Figure 1 shows only the functional blocks relevant to this embodiment. The network switch 1 comprises a receiving unit 2, a routing unit 3, a transmission control unit 4, and a transmitting unit 5.

The receiving unit 2 has N input ports 21 (N is an integer greater than or equal to 1). In Figure 1, the N input ports 21 are referred to as input ports 21-1 to 21-N. The input ports 21 are input interfaces that input packets from the outside. Each input port 21 of the receiving unit 2 receives packets from the outside and outputs the received packets to the routing unit 3. The transmitting unit 5 has M output ports 51 (M is an integer greater than or equal to 1). In Figure 1, the M output ports 51 are referred to as output ports 51-1 to 51-M. The output ports 51 are output interfaces that output packets to the outside. The output ports 51 input packets from the transmission control unit 4 and output the input packets to a transmission line between the output ports 51 and an external device.

The routing unit 3 determines the output port 51 to output the packet using a process similar to any conventional technology. The routing unit 3 has a routing information storage unit 31 and a routing processing unit 32. The routing information storage unit 31 stores routing information. The routing information indicates the correspondence between address information and output ports 51. The routing processing unit 32 reads the output port 51 corresponding to the address information set in the packet from the routing information as the output destination. Address information such as a MAC address or IP address is used. Other information in addition to address information may be used to determine the output destination. Furthermore, if the address information of the packet indicates a multicast address, the routing processing unit 32 duplicates the packet and sets each duplicated packet as the output destination for each output port 51. The routing processing unit 32 outputs the packet and output destination information indicating the output port 51 to the transmission control unit 4. Note that a packet whose output destination is output port 51-m (m is an integer between 1 and M) is referred to as a packet of output port 51-m.

The transmission control unit 4 comprises a packet type determination unit 41, a flow information measurement unit 42, a flow status storage unit 43, an output control unit 44, and a queue unit 45. The packet type determination unit 41 analyzes the packet header of the packet input from the routing processing unit 32 and reads the flow identification information and the RTP packet marker bit (hereinafter referred to as the "M bit"). The flow identification information is information that identifies an IP flow. The flow identification information includes the source IP address, destination IP address, L4 protocol type, source L4 port, and destination L4 port. The M bit indicates whether the packet contains the last byte of the video field. In progressive scan, a video field corresponds to one frame of video. In other words, in progressive scan, the M bit of a packet containing the last byte of one frame in the RTP payload is set to 1. In video interlacing, one frame of video is composed of two video fields. In video interlacing, the M bit of a packet containing the last byte of each video field in the RTP payload is set to 1. The M bit of other packets is set to 0. The packet type determination unit 41 outputs the packet, flow identification information, and M bit to the flow information measurement unit 42.

The flow information measurement unit 42 measures each measurement item of the IP flow. Based on the measurement results, the flow information measurement unit 42 updates the flow state information and thinning table stored in the flow state storage unit 43. The flow state storage unit 43 stores flow state information for each IP flow in a temporary storage area such as a memory. The flow state storage unit 43 also stores a thinning table for each output port 51.

Figure 2 shows an example of flow state information. The flow state information includes output port information for the IP flow, flow identification information, cumulative transfer data volume, packet count, packet count when the previous M bit was 1, M-bit count, number of packets between M bits, input bit rate, input frame rate, thinning rate, frame count during thinning, QoS value, measurement completion flag, last received timestamp, and thinning pattern. The output port information and flow identification information are key information used to identify the flow state information.

The output port information indicates the output port 51 from which packets of an IP flow are output. The cumulative transfer data volume is the total data volume from the first packet of the IP flow received by the network switch 1 to the newest packet. The packet count is the number of packets received by the network switch 1 from the first packet of the IP flow to the newest packet. The M-bit count is the number of packets with an M-bit of 1 in the IP flow received by the network switch 1. The inter-M-bit packet count is the number of packets received by the network switch 1 from a packet with an M-bit of 1 to the next packet with an M-bit of 1. The thinning rate indicates the rate at which frames are discarded when the network switch 1 is thinning out. Thinning out refers to discarding some of the received packets and not outputting them externally from the network switch 1. The frame count during thinning out is the number of frames received by the network switch 1 while thinning out. The QoS value is a value representing priority. In this embodiment, there are two types of priority: priority and normal. The measurement flag indicates whether the input bit rate and input frame rate have been measured. The last received timestamp is the time when the network switch 1 last received a packet of the IP flow. The thinning pattern indicates the priority of each frame received during thinning.

Figure 3 shows an example of a thinning table. The thinning table has an entry for each output port 51. The thinning table contains information that associates the flow identification information of each IP flow, the thinning pattern, the input FPS (frames per second), the input bit rate, the output FPS, and the output bit rate.

The thinning patterns in the thinning table are the same as the thinning patterns in the flow status information and are updated synchronously. In Figure 3, the priority of each frame being thinned out is represented using "1" and "0." "1" is the symbol representing priority, and "0" is the symbol representing normal. Numbers, letters, and symbols other than "1" and "0" can also be used as symbols. The position of the symbol from the left in the thinning pattern represents the order of the frames being thinned out. In other words, the thinning pattern "00000011..." indicates that the first through sixth video frames during the thinning period have normal priority, while the seventh and eighth frames have priority. In Figure 3(a), the thinning patterns for all IP flows include periods in which no frames are thinned out. Therefore, the thinning patterns in the thinning table shown in Figure 3(a) can be removed from the periods in which no frames are thinned out, resulting in the pattern shown in Figure 3(b).

The input FPS indicates the frame rate of the IP flow input to the network switch 1. The input bit rate indicates the bit rate of the IP flow input to the network switch 1. The output FPS indicates the frame rate of the IP flow output from the network switch 1. The output bit rate indicates the bit rate of the IP flow output from the network switch 1.

The output control unit 44 shown in FIG. 1 determines the priority of packets and outputs a QoS value representing the determined priority to the flow information measurement unit 42. If the bandwidth of data to be output from the destination output port 51-m (m is an integer between 1 and M) of the packets is equal to or less than a threshold, the output control unit 44 determines that all packets from output port 51-m are prioritized. The threshold is a value equal to or less than the interface speed of output port 51-m. On the other hand, if the bandwidth of data to be output from output port 51-m exceeds the threshold, the output control unit 44 determines that congestion has occurred and performs thinning. When performing thinning, the output control unit 44 determines the priority of packets from output port 51-m according to a thinning table. The output control unit 44 outputs a QoS value representing the determined priority to the flow information measurement unit 42. The flow information measurement unit 42 places the packets in a transmission queue corresponding to the QoS value.

The queue unit 45 has transmission queues corresponding to the QoS values. In this embodiment, the queue unit 45 has two FIFO (First in First out) transmission queues: a priority transmission queue Q1 and a normal transmission queue Q2. The priority transmission queue Q1 temporarily stores packets with a priority priority, while the normal transmission queue Q2 temporarily stores packets with a normal priority. The queue unit 45 reads packets from the priority transmission queue Q1 and outputs them to the output port 51 indicated by the output destination information associated with the read packets. If there are no packets stored in the priority transmission queue Q1, the queue unit 45 reads packets from the normal transmission queue Q2 and outputs them to the output port 51 indicated by the output destination information associated with the read packets. The queue unit 45 also discards packets that have been stored in the normal transmission queue Q2 for more than a predetermined time.

Next, the operation of the network switch 1 will be described.
4 is a flow diagram showing the transfer control process of the network switch 1. The network switch 1 performs the process of FIG. 4 every time it receives a packet. The input port 21 of the network switch 1 receives the packet. The received packet is referred to as a received packet. The input port 21 outputs the received packet to the routing unit 3 (step S101).

The routing unit 3 determines the output interface to output the received packet (step S102). Specifically, the routing processing unit 32 reads address information from the received packet. From the routing information stored in the routing information storage unit 31, the routing processing unit 32 reads the output port 51-m corresponding to the address information as the output destination. Hereinafter, the output port 51-m to which the received packet is output will be referred to as the transmission output port 51-m. The routing processing unit 32 adds destination information indicating the transmission output port 51-m to the received packet and outputs it to the packet type determination unit 41. Note that if the address information indicates a multicast address, the routing processing unit 32 duplicates the received packet and sets the output destination of each of the M duplicated received packets to output ports 51-1 to 51-M. In this case, the network switch 1 performs the processes from step S103 onwards for each duplicated received packet.

The packet type determination unit 41 analyzes the packet header of the received packet input from the routing processing unit 32 and reads the source IP address, destination IP address, L4 protocol type, source L4 port, destination L4 port, and M bit (step S103). The packet type determination unit 41 adds flow identification information containing the read source IP address, destination IP address, L4 protocol type, source L4 port, and destination L4 port, as well as the M bit, to the received packet and outputs it to the flow information measurement unit 42.

The flow information measurement unit 42 determines whether the received packet is UDP/RTP (step S104). If the flow information measurement unit 42 determines that the packet is UDP/RTP (step S104: YES), it identifies a flow state information entry stored in the flow state storage unit 43 based on the combination of the transmission output port 51-m and the flow identification information (step S105). If there is no entry, the flow information measurement unit 42 creates a flow state information entry corresponding to the combination of the transmission output port 51-m and the flow identification information in the flow state storage unit 43. The flow information measurement unit 42 sets the output port information and flow identification information in the created flow state information, and sets NULL for other information. The output port information can be output destination information. In the processing from step S106 onwards, the transmission control unit 4 processes the flow state information identified or created in step S105.

The flow information measurement unit 42 updates the cumulative transfer data volume set in the flow state information to a value obtained by adding the payload size of the received packet (step S106). The flow information measurement unit 42 determines whether the M bit of the received packet is 1 (step S107). If the flow information measurement unit 42 determines that the M bit is 0 (step S107: NO), it updates the packet count number set in the flow state information to a value obtained by adding 1 (step S108). The flow information measurement unit 42 updates the last reception timestamp set in the flow state information to a value representing the current time (step S109).

The output control unit 44 determines the QoS value of the received packet (step S110). The determination of the QoS value will be explained later with reference to Figure 6. The output control unit 44 outputs the determined QoS value to the flow information measurement unit 42. If the QoS value indicates priority, the flow information measurement unit 42 stores the received packet in the priority transmission queue Q1. If the QoS value indicates normal, the flow information measurement unit 42 stores the received packet in the normal transmission queue Q2 (step S111). If a packet is stored in the priority transmission queue Q1, the queue unit 45 reads the packet from the priority transmission queue Q1 and outputs it to the output port 51 indicated by the destination information attached to the read packet. If a packet is not stored in the priority transmission queue Q1, the queue unit 45 reads the packet from the normal transmission queue Q2 and outputs it to the output port 51 indicated by the destination information attached to the read packet. The output port 51 outputs the packet input from the queue unit 45 to the transmission line (step S112). The queue unit 45 discards packets that have been stored in the normal transmission queue Q2 for a predetermined period of time or longer. This causes at least some of the packets with normal priority to be discarded. Alternatively, the queue unit 45 may discard all packets stored in the normal transmission queue Q2.

On the other hand, if the flow information measurement unit 42 determines in step S107 that the M bit is 1 (step S107: YES), it updates the M-bit count number set in the flow state information by adding 1 (step S113). The flow information measurement unit 42 reads the packet count number and the packet count number when the previous M bit was 1 from the flow state information. The flow information measurement unit 42 subtracts the packet count number when the previous M bit was 1 from the packet count number to calculate the number of packets for one frame. The flow information measurement unit 42 updates the M-bit packet count number set in the flow state information to the calculated number of packets for one frame (step S114). Furthermore, the flow information measurement unit 42 updates the packet count number when the previous M bit was 1 set in the flow state information to the current packet count number set in the flow state information (step S115).

The flow information measurement unit 42 determines whether the transmission output port 51-m is currently thinning out (step S116). If the flow information measurement unit 42 determines that thinning out is not in progress (step S116: NO), it executes the process from step S108. If the flow information measurement unit 42 determines that thinning out is in progress (step S116: YES), it updates the thinning-out frame count number set in the flow status information by adding 1 (step S117).

The flow information measurement unit 42 identifies the thinning table for the transmission output port 51-m and obtains the number of symbols set in the thinning pattern as the number of frames. The flow information measurement unit 42 reads the frame count during thinning and the thinning pattern from the flow state information. The flow information measurement unit 42 determines whether the frame count during thinning exceeds the number of frames represented by the thinning pattern (step S118). If the flow information measurement unit 42 determines that the number of frames represented by the thinning pattern has not been exceeded (step S118: NO), it executes the processing from step S108. If the flow information measurement unit 42 determines that the number of frames represented by the thinning pattern has been exceeded (step S118: YES), it sets the frame count during thinning in the flow state information to 0 (step S119), and then executes the processing from step S108.

If the flow information measurement unit 42 determines that the received packet is not UDP/RTP (step S104: NO), the output control unit 44 determines a QoS value with a predetermined priority (step S120). The network switch 1 performs processing from step S111.

Figure 5 is a flow diagram showing the transfer rate measurement process of the network switch 1. The flow information measurement unit 42 has an internal interrupt timer that operates, for example, every second. When an interrupt occurs, the flow information measurement unit 42 starts the process of Figure 5 for each piece of flow state information. The flow information measurement unit 42 may perform the process of Figure 5 only for flow state information whose measurement completion flag is 0.

The flow information measurement unit 42 calculates the amount of change from one second ago in the integrated transfer volume set in the flow state information. The flow information measurement unit 42 sets the calculated amount of change as the input bit rate of the flow state information (step S201). If there is a thinning table identified by the output port 51 and flow identification information set in the flow state information, the flow information measurement unit 42 sets the calculated bit rate as the input bit rate associated with the flow identification information in the thinning table.

Furthermore, the flow information measurement unit 42 calculates the amount of change in the M bits set in the flow state information from one second ago. The flow information measurement unit 42 updates the input frame rate of the flow state information based on the calculated amount of change (step S202). For example, if the amount of change in M bits per second is 24, this indicates that 24 frames of packets were transferred in one second. Therefore, the flow information measurement unit 42 sets the input bit rate to 24 fps. Similarly, if the amount of change is 30, the flow information measurement unit 42 sets the input frame rate of the flow state information to 30 fps, and if the amount of change is 60, it sets the input frame rate of the flow state information to 60 fps. If there is a thinning table identified by the output port 51 and flow identification information set in the flow state information, the flow information measurement unit 42 sets the input frame rate based on the calculated amount of change to the input FPS associated with the flow identification information in the thinning table.

The flow information measurement unit 42 sets a measurement completion flag in the flow status information (step S203). This is because the same input bit rate and input frame rate continue during video distribution.

The flow information measurement unit 42 resets the measurement completion flag when routing information is changed in response to a request from a receiving device sent using a protocol such as IGMP or PIM (Protocol Independent Multicast).

Figure 6 is a flow diagram showing the QoS value determination process of the network switch 1. Figure 6 shows detailed processing of step S110 in Figure 4. The output control unit 44 identifies the flow state information of all IP flows whose destination is the transmission output port 51-m indicated by the output destination information of the received packet. That is, the output control unit 44 identifies the flow state information for which the transmission output port 51-m is set in the output port information. The output control unit 44 calculates the sum of the input bit rate values set in each identified flow state information (step S301). The calculated sum represents the bandwidth of the data to be output from the transmission output port 51-m. Note that the output control unit 44 may obtain the input bit rate values of each IP flow from the thinning table of the entries for the transmission output port 51-m and calculate their sum. An IP flow whose packet output destination is the transmission output port 51-m is referred to as the IP flow of the transmission output port 51-m.

The output control unit 44 determines whether the sum of the input bit rate values is less than a threshold value (step S302). The threshold value is the interface speed of the transmission output port 51-m. The threshold value may be a value obtained by multiplying the interface speed of the transmission output port 51-m by a coefficient less than 1, or a predetermined value that does not exceed the interface speed of the transmission output port 51-m. If the output control unit 44 determines that the sum of the input bit rate values is less than the threshold value (step S302: YES), it determines that all IP flow packets of the transmission output port 51-m are priority. In other words, the output control unit 44 determines the QoS value of the received packet to a value that represents priority (step S303). The output control unit 44 sets the determined QoS value in the flow status information of the received packet. Furthermore, if packets of the transmission output port 51-m are currently being thinned out, the output control unit 44 ends the thinning.

If the output control unit 44 determines that the sum of the input bit rate values exceeds the threshold (step S302: NO), it determines whether packets are currently being thinned out at the transmission output port 51-m (step S304). If the output control unit 44 determines that packets are not being thinned out (step S304: NO), it enables the thinning output operation of the transmission output port 51-m (step S305). During the thinning output operation, the network switch 1 assigns received packets to a priority transmission queue or a normal transmission queue based on the thinning rate, thinning-out frame count, thinning pattern, and measurement completion flag set in the flow state information.

The output control unit 44 sets the thinning-out frame count in the flow state information of the received packet to an initial value of 1. Furthermore, the output control unit 44 sets the thinning-out frame count in the flow state information of each IP flow of the transmission output port 51-m to an initial value of 0, except for the flow state information of the received packet (step S306). The output control unit 44 updates the thinning rate for all IP flows of the transmission output port 51-m (step S307). Therefore, the output control unit 44 calculates the thinning rate for the transmission output port 51-m using the following formula (1):

Thinning rate = 1 - (output interface bandwidth / input bitrate) ... (1)

The bandwidth of the output interface is the bandwidth that the transmission output port 51-m can output, and can be, for example, the threshold value used in step S302. The output control unit 44 calculates the input bit rate as the sum of the input bit rates read from the flow status information for all IP flows of the transmission output port 51-m, or the sum of the input bit rates for all IP flows set in the thinning table for the transmission output port 51-m. The output control unit 44 updates the thinning rate of the flow status information for all IP flows of the transmission output port 51-m to the thinning rate calculated using equation (1) (step S307).

The output control unit 44 then updates the output frame rate and output bit rate for each IP flow of the transmission output port 51-m to the values after thinning. The output control unit 44 calculates the output frame rate and output bit rate for each IP flow of the transmission output port 51-m using the following equations (2) and (3), respectively.

Output frame rate = IP flow input frame rate × thinning rate … (2)

Output bit rate = IP flow input bit rate × thinning rate … (3)

The output control unit 44 reads the input frame rate and input bit rate of each IP flow of the transmission output port 51-m from the flow status information of each IP flow. Alternatively, the output control unit 44 may read the input FPS and input bit rate of each IP flow from the thinning table of the transmission output port 51-m. The output control unit 44 updates the output FPS and output bit rate of each IP flow set in the thinning table of the transmission output port 51-m to the output frame rate and output bit rate calculated for each IP flow (step S308).

The output control unit 44 sets a thinning pattern for each IP flow set in the thinning table of the transmission output port 51-m (step S309). The output control unit 44 sets the number of input FPSs for each IP flow set in the thinning table of the transmission output port 51-m as the number of symbols p of the thinning pattern. The output control unit 44 sets the symbol "1" in the thinning pattern, indicating that frames equal to the number of output FPSs q will be output with priority, and sets the symbol "0" for the remaining (p-q) frames, indicating that they are to be thinned. At this time, the output control unit 44 sets (p-q) consecutive "0s" in order from the lowest number in the flow identification information, so that the time intervals of frames set to 0 differ between flows. This ensures that the timing at which frames are output with priority and the timing at which they are discarded are distributed between IP flows.

Figure 3(a) shows an example in which four IP flows, F1 to F4, with an input bit rate of 3 Gbps, are output to a 10 Gbps (gigabits per second) output interface. The sum of the input bit rates is 12 Gbps. From equation (1), the thinning rate is 1 - (10 Gbps / 12 Gbps) = 0.83.... The output frame rate of each of flows F1 to F4 is 30 fps. From equation (2), the output frame rate of each of flows F1 to F4 after thinning is 0.8333 x 30 = 24.999 fps, which is rounded down to 24 fps.

Since the input FPS is 30 fps, the number of symbols set in the thinning pattern is 30. Since the output frame rate is 24 fps, the output control unit 44 sets 24 of the 30 symbols to 1 and 6 consecutive symbols to 0. The output control unit 44 sets the thinning patterns so that the positions where 0s are set in the thinning patterns for flows F1 to F4 do not overlap, and so that 0s are closer to the beginning across flows F1 to F4.

In this way, the thinning table is set as shown in Figure 3(a). In the thinning table in Figure 3(a), the thinning patterns for the latter six frames are all "1". Therefore, the output control unit 44 terminates the thinning pattern at the output frame rate of 24 fps, resulting in the thinning table shown in Figure 3(b). In this way, the output control unit 44 deletes frames for which all IP flows have a thinning pattern of "1" from the thinning table.

If the output control unit 44 determines in step S304 that thinning is in progress (step S304: YES), or after processing step S310, the output control unit 44 performs processing step S311. That is, the output control unit 44 reads the thinning-in-progress count from the flow state information of the received packet. Furthermore, the output control unit 44 reads from the thinning table a thinning pattern corresponding to the flow identification information of the received packet, and reads from the read thinning pattern a symbol at a position corresponding to the thinning-in-progress count. The output control unit 44 determines whether the read symbol is 1 (step S311). If the output control unit 44 determines that the symbol is 1 (step S311: YES), it determines the QoS value of the received packet to a value representing priority (step S312). If the output control unit 44 determines that the symbol is 0 (step S311: NO), it determines the QoS value of the received packet to a value representing normal (step S313). As a result, in the case of the thinning table shown in Figure 3(b), six frames of received packets for each of all IP flows at transmission output port 51-m are either placed in the normal transmission queue or discarded.

FIG. 7 is a flow diagram showing the entry deletion process of the network switch 1. The flow information measurement unit 42 executes the process shown in FIG. 7 at predetermined time intervals. The flow information measurement unit 42 selects all entries in the flow state storage unit 43 one by one (step S401). An entry is a combination of an output port 51 and a flow identifier. The flow information measurement unit 42 identifies the flow state information of the selected entry. The flow information measurement unit 42 determines whether the last reception timestamp of the identified flow state information has been updated since the previous process of FIG. 7 was performed (step S402). If the flow information measurement unit 42 determines that an update has occurred (step S402: YES), it selects the next entry. If the flow information measurement unit 42 determines that no update has occurred (step S402: NO), it deletes the identified flow state information and then selects the next entry (step S403). If all entries have been selected, the flow information measurement unit 42 terminates the process of FIG. 7.

In the above example, the thinning pattern was set so that 0's are consecutive. Setting it in this way makes it easier to set the thinning pattern when the input frame rates of each IP flow are different. When the input frame rate of all IP flows is the same, the 0's do not need to be consecutive as long as they are distributed among the IP flows. For example, in the thinning table of Figure 3(b), the thinning pattern for flow F1 may be "10001000...1000", the thinning pattern for flow F2 may be "01000100...0100", the thinning pattern for flow F3 may be "00100010...0010", and the thinning pattern for flow F4 may be "00010001...0001".

The network switch 1 in the above-described embodiment may include a CPU (Central Processing Unit), memory, auxiliary storage device, etc., connected via a bus, and the routing unit 3 and transmission control unit 4 may be implemented by the CPU executing a program. Note that all or part of the functions of the routing unit 3 and transmission control unit 4 may be implemented using hardware such as an ASIC (Application Specific Integrated Circuit), a PLD (Programmable Logic Device), or an FPGA (Field Programmable Gate Array). The program may be recorded on a computer-readable recording medium. Examples of computer-readable recording media include portable media such as flexible disks, optical magnetic disks, ROMs, and CD-ROMs, as well as storage devices such as hard disks built into computer systems. The program may also be transmitted via a telecommunications line.

According to the above embodiment, if the video packets output by the network switch 1 exceed the transfer rate, the QoS value is dynamically set for each video frame, suppressing the output rate while allowing the receiving device of the video packets to construct video frames, albeit intermittently. In this way, even if congestion occurs in the network switch 1, it is possible to reduce long periods of video interruption at the receiving device.

In the embodiment described above, the packet forwarding device includes a receiving unit, a transmitting unit, an output control unit, and a priority control unit. For example, the packet forwarding device is the network switch 1 of the embodiment. The receiving unit receives packets containing data constituting each frame of video. The output unit outputs the packets received by the receiving unit. The output control unit determines the priority of packets on a frame-by-frame basis when the bandwidth of packets output from the output unit exceeds a threshold. The priority control unit discards at least some of the packets received by the receiving unit that contain data of frames assigned a low priority by the output control unit, and outputs the packets that were not discarded to the output unit. The priority control unit is, for example, the queue unit 45 of the embodiment.

The packet forwarding device may further include a determination unit and a measurement unit. The determination unit determines the flow based on address information set in the packets. For example, the determination unit is the packet type determination unit 41 in the embodiment. The measurement unit measures the input speed of packets to the receiving unit. For example, the measurement unit is the flow information measurement unit 42 in the embodiment. The output control unit calculates a thinning rate, which is the rate at which frames are thinned out, based on the input speed measured by the measurement unit and a threshold value. The output control unit assigns a low priority to some frames for each flow according to the calculated thinning rate, so that the periods during which low priority is assigned are distributed among flows.

The output unit may have multiple output interfaces that output packets to the line. The measurement unit measures the input speed of packets to the receiving unit for each flow. The output control unit calculates a thinning rate for each output interface based on the sum of the input speeds of each flow that outputs packets from the output interface and a threshold value corresponding to the output interface, and assigns a low priority to some frames for each flow according to the calculated thinning rate so that the periods during which low priority is assigned are distributed among the flows that output packets from the output interface.

The above describes in detail an embodiment of the present invention with reference to the drawings, but the specific configuration is not limited to this embodiment and includes designs that do not deviate from the gist of the present invention.

1 Network switch 2 Receiving unit 3 Routing unit 4 Transmission control unit 5 Transmitting unit 21 Input ports 21-1 to 21-N Input port 21-N Input port 31 Routing information storage unit 32 Routing processing unit 41 Packet type determination unit 42 Flow information measurement unit 43 Flow state storage unit 44 Output control unit 45 Queue unit 51-1 to 51-M Output port Q1 Priority transmission queue Q2 Normal transmission queue

Claims (3)

a receiving unit that receives packets containing data that constitute each frame of video;
an output unit that outputs the packet received by the receiving unit;
an output control unit that determines a priority of the packet on a frame-by-frame basis when a bandwidth of the packet output from the output unit exceeds a threshold;
a priority control unit that discards at least a portion of packets that have data of the frame to which a low priority is assigned by the output control unit, among the packets received by the receiving unit, and outputs the packets that have not been discarded to the output unit;
a determination unit that determines a flow based on address information set in the packet;
a measuring unit that measures an input speed of the packets to the receiving unit;
Equipped with
the output control unit calculates a thinning rate, which is a rate at which the frames are thinned out, based on the input speed measured by the measurement unit and the threshold, and assigns a low priority to some of the frames for each flow according to the thinning rate so that periods during which low priorities are assigned are dispersed among the flows.
Packet forwarding equipment. the output unit has a plurality of output interfaces for outputting the packets to a line;
the measuring unit measures an input speed of the packets to the receiving unit for each flow,
the output control unit performs a process of calculating a thinning rate for each of the output interfaces based on the sum of the input rates of the flows that output the packets from the output interface and a threshold value according to the output interface, and a process of assigning a low priority to some of the frames for each of the flows according to the thinning rate so that periods during which a low priority is assigned are distributed among the flows that output the packets from the output interface.
2. The packet forwarding device according to claim 1 . On the computer,
an output control step of determining a priority of a packet on a frame-by-frame basis when a bandwidth of the packet received by the receiving unit and output from an output unit that outputs the packet and sets data constituting each frame of video exceeds a threshold;
a priority control step of discarding at least a part of the packets received by the receiving unit, in which data of the frame to which a low priority is assigned in the output control step is set, and controlling so that the packets that are not discarded are output to the output unit;
a determination step of determining a flow based on address information set in the packet;
a measuring step of measuring an input speed of the packets to the receiving unit;
Execute
In the output control step, a thinning rate is calculated based on the input speed measured in the measurement step and the threshold value, and a low priority is assigned to some of the frames for each flow in accordance with the thinning rate so that periods during which low priority is assigned are dispersed among the flows.
Program for.