RETRANSMISSION OF DATA IN PACKET NETWORKS

Abstract

A method in a packet-based network like the Internet is provided, which is directed towards packet loss recovery for transmission of a data stream, DS.sub.0 in a packet-based network. The method comprises transmitting data packets associated with the data stream, and upon receiving a retransmission request of missing data, retransmitting the missing data applying a selective retransmission of missing data, which may be based on a priority ranking of data types to be retransmitted, performing retransmission if it is determined that missing data will be received in time for display, or performing retransmission with a rate control.

Claims

1. A method for providing packet loss recovery for transmission of a data stream in a packet-based network comprising: transmitting data packets associated with said data stream; and upon receiving a retransmission request of missing data; retransmitting said missing data; and applying a selective retransmission of missing data.

2. A method according to claim 1, wherein said selective retransmission is based on a priority ranking of data types to be retransmitted.

3. A method according to claim 1, wherein said selective retransmission comprises at least one of performing retransmission if it is determined that missing data will be received within a predetermined time period, and performing retransmission with a rate control.

4. A method according to claim 1, further comprising monitoring characteristics of the network comprising at least one of quality of service, QoS, transmission capacity or delay of the network, and wherein retransmission is activated based on said monitored characteristics.

5. A method according to claim 1, wherein said transmitting of data packets comprises: dividing each data packet of said data stream into a preselected number of fragments; generating macro-packets in which said fragments of said data packet are distributed; and transmitting said macro-packets.

6. A method according to claim 5, wherein said missing data corresponds to at least one missing fragment, and wherein said retransmitting of said missing data comprises: distributing said missing fragments into currently generated macro-packets, and retransmitting said missing fragments.

7. A method according to claim 5, wherein said generated macro-packets comprise fragments from multiple data packets.

8. A method according to claim 1, further comprising adding FEC fragments for said missing data.

9. A method for providing packet loss recovery for transmission of a data stream in a packet-based network comprising: dividing each data packet of said data stream into a preselected number of fragments, and generating macro-packets in which said fragments of said data packet are distributed; transmitting said macro-packets; and upon receiving a retransmission request of missing fragments of a data packet: distributing said missing fragments into currently generated macro-packets, and retransmitting said missing fragments.

10. A method according to claim 9, wherein said generated macro-packets comprise fragments from multiple data packets.

11. A method according to claim 9, further comprising adding FEC fragments for said missing fragments.

12. A method according to any of claim 9, further comprising applying selective retransmission of missing fragments.

13. A method according to claim 12, wherein said selective retransmission is based on at least one of a priority ranking of data types to be retransmitted, performing retransmission if it is determined that missing fragments will be received within a predetermined time period, and performing retransmission with a rate control.

14. A method according to claim 9, further comprising monitoring characteristics of the network comprising at least one of QoS, transmission capacity or delay of the network, and wherein said selective retransmission is activated based on said monitored characteristics.

15. A method according to claim 2 or 13, wherein said data types comprise at least one of audio data, complete video image data, and delta changes of images data.

16. A node in a communication network comprising means for performing a method according to claim 9.

17. A software module adapted to perform a method according to claim 1 when executed by a computer processor.

Description

DRAWINGS

[0021] The above, as well as additional objects, features and advantages of the present invention, will be better understood through the following illustrative and non-limiting detailed description of preferred embodiments of the present invention, with reference to the appended drawings, where the same reference numerals will be used for similar elements, wherein:

[0022] FIG. 1 is a schematic block diagram illustrating a distribution system in which embodiments according to the present inventive concept may be employed;

[0023] FIG. 2 is a flowchart illustrating an embodiment of a method according to the present inventive concept;

[0024] FIG. 3 is a schematic block diagram illustrating an embodiment of the inventive concept;

[0025] FIGS. 4 to 6 are flowcharts illustrating embodiments of a method according to the present inventive concept; and

[0026] FIGS. 7a and 7b are schematic block diagrams illustrating an embodiment of the inventive concept.

[0027] All the figures are schematic, not necessarily to scale, and generally only show parts which are necessary in order to elucidate the invention, wherein other parts may be omitted or merely suggested.

DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS

[0028] FIG. 1 is a block diagram schematically illustrating a distribution network system 10 of IP type for live distribution of e.g. video, in view of which aspects of the inventive concept will be described. The distribution system 10 includes a server 11 (a source) and one or more recipients or client devices 15, 16, 17 coupled by a respective communication link over a network 12. In the distribution network system 10 data transmission of a data stream DSc, from the server 11 to one or more of the client devices 15, 16, 17 over the network 12 comprises steps according to embodiments of the present invention. The data transmission may involve transmitting e.g. video content or other media content in the form of video packets (multi cast video packets). In order to handle missing packets and other types of errors, communication and distribution systems employ various techniques to handle erroneously received information. The client device may correct the erroneously received information amongst other techniques by retransmission techniques, which enable the erroneously received information to be retransmitted to the receiver, for example, by using automatic retransmission request (ARQ) or forward error correction (FEC) techniques. FEC techniques include, for example, convolutional or block coding of the data prior to modulation. FEC coding involves representing a certain number of data bits or blocks of data using a certain (greater) number of code bits or code blocks, thereby adding redundancy which permits correction of certain errors.

[0029] If there are lost packets during data transmission, any client(s) experiencing missing packets will request retransmission of the lost packets from the server 11. In addition to retransmission if there is FEC information available, the clients can either attempt to recover lost packets by FEC correction and then by retransmission, or first by retransmission and then by FEC correction on the newly transmitted data.

[0030] The communication link may be a computer network (e.g. a LAN a WAN, the Internet), a wireless network (e.g. a cellular data network), or some combination of these network types.

[0031] A flowchart illustrating steps of an embodiment of a method for providing packet loss recovery for transmission of a data stream DS.sub.0 in a packet-based network is presented in FIG. 2 and described with further reference to FIGS. 1 and 3. For illustrative purposes, the data stream for transmission DS.sub.0 is represented as a sequence of data packets of a contiguous stream of data, with each data packet comprising a set of payload information corresponding to a specific data type.

[0032] According to a simplified exemplifying embodiment as described below with reference to FIG. 3, the data types comprise at least one of audio data (A), complete video image/full picture data (FP), and delta changes of images data (S). In a first step (step S120, see FIGS. 2 and 3), data packets of data stream DS.sub.0 are transmitted as data stream DS to a client device, e.g. client device 16. The server 11 is arranged to continuously monitor if retransmission requests of missing data of an already transmitted data packet is received from the client device (step S130). If a retransmission request REQ of missing data (or a data packet) is received (step S140), selective retransmission is applied to the missing data (S145). The selective retransmission mechanism may be applied at the client device 16, where based on a current traffic situation, available resources, or capacity of the client, selective retransmission requests are made. For instance, if the client device determine that the quality of its connection with the server is one of excellent, sufficient or poor (e.g. via 4G or via a low quality Wi-Fi signal), it requests retransmission of all or only part of the missing data, respectively.

[0033] Alternatively, the selective retransmission is determined based on information at the server 11. Since the server 11 is typically connected to multiple client devices, and other nodes of the network, decisions of the selective retransmission may be determined on a network level.

[0034] The missing data accepted for selective retransmission is interleaved in the outgoing data stream DS (step S150), and thus retransmitted to the client device.

[0035] In an optional step S148 (additional step with dashed lines in FIG. 2), the method further comprises adding FEC packets for the missing data. The FEC packets are preferably distributed into the outgoing data stream DS in the same manner as missing data for retransmission. Advantageously, if a resent data packet is dropped, the data packet can be reconstructed by means of corresponding FEC packets, and additional retransmission of the missing data packet may be avoided.

[0036] The retransmission distribution of missing data may be arranged according to a preselected distribution scheme where the requested data is distributed e.g. in every n.sup.th transmitted packet (performing retransmission with a rate control), or not retransmitted governed by time limitations. For instance, when the data stream is part of a live broadcast such as live TV distribution, retransmission of a missing data may be down prioritized if the expected arrival time at the client is too late for playout. The selective retransmission is according to embodiments based on a priority ranking of data types to be retransmitted, or priority ranking based on number of retransmissions of certain data etc.

[0037] In an exemplifying embodiment as illustrated in FIG. 3, the method as described with reference to FIG. 2 will now be visualized starting with a data stream DS.sub.0 for transmission which comprises a number of subsequent packets 1-N (each illustrated as a rectangle). Considering the data stream to be an MPEG transport stream MPEG-TS, retransmission in the IP network is typically performed on an application level or at the MPEG-TS level utilizing a user-defined-retransmission scheme. In this simplified illustration each packet represents 188 bytes MPEG packets which each contains audio samples (A), different types of reference pictures/frames (e.g. Instantaneous Decoding Refresh, IDR, frames or I-frames, B-frames) indicated as full picture data (FP) and inter frame data (B-frames, P-frames) here indicated as sub picture data (S). IDR frames are typically larger than a MPEG packet and are thus transported in multiple consecutive MPEG packets. In a packet header, the type of data contained in the corresponding packet (and coming packets) is identified. For MPEG packets containing IDR frames, the first header identifies that IDR data is coming in consecutive packets, followed by e.g. packets (each) containing a number of audio samples etc., each packet having a given sequence number. As illustrated in FIG. 3, at the sending side, subsequent data packets with sequence numbers 1, 2, 3, 4, . . . 9, 10 . . . N of the data stream DS.sub.0 are transmitted as data stream DS (S120) which at the client side is received as data stream DS.sub.2. At reception of the data stream DS.sub.2, it is determined that data packets with sequence numbers 3-5 are missing and the client device in this example then sends a request REQ for the all missing data packets 3-5 (S130) to the sending side. Based on the request REQ and, according to embodiments of the method, on additional information regarding the current state of the network, e.g. if the client device is connected to 4G or Wi-Fi, selective retransmission is applied. In the illustration selective retransmission based on data type prioritization is utilized. The priority order is selected from highest to lowest priority in the following order: audio data A, full picture data FP, and sub picture data S. To limit the load on the network, here only full picture data FP and audio data A (i.e. data packets 4 and 5) are retransmitted. The missing packets 4 and 5 are interleaved in the currently transmitted data stream DS (S146).

[0038] Continuing now with reference to FIG. 1 and to FIG. 4, which is a flowchart illustrating steps in an embodiment of a method for providing packet loss recovery for transmission of a data stream DS.sub.0 in a packet-based network according to the present inventive concept. For illustrative purposes, the data stream for transmission DS.sub.0 is represented as a sequence of data packets representing a contiguous stream of information, with each data packet comprising a set of payload information representative of a segment of the stream of information corresponding thereto. In the exemplifying embodiments subsequent packets are denoted as [P(X+1)], [P(X)], . . . , [P(2)], [P(1)], the number of segments for dividing the packets N is selected to be 4, and the size of macro-packets and size of the data packets are selected to be the same. Preferably, the size of the macro-packets and the size of the data packets are selected as close as possible to the maximum size of an Ethernet packet (which is 1518 or 1522 bytes). In the case of using MPEG-TS mapping, the fragments are MPEG-TS packets which have a size of 188 bytes (or if the optional trailer is used 208 bytes) meaning that there is 7 fragments in an IP/Ethernet packet. The invention is not limited to MPEG-TS code stream, any digital media flow may be protected using the invented algorithm such as MPEG-TS on RTP or directly on RTP. The selected number of segments N and the size of the macro-packets are application dependent and preferably selected depending on the data transmission standard used in the distribution.

[0039] Before transmission of the data stream from the server 11, the information payload of each data packet is divided into N fragments (step S100). The fragments are normally constructed in an encoder or a transcoder. Each fragment has a header identifying what type of data the fragment has, etc. Packet [P(1)] is thus divided into [P(1):4], [P(1):3], [P(1): 2], [P(1):1]. Next, macro-packets MP.sub.n are generated in which fragments of a predetermined number of packets are distributed (step S110), when a macro-packet is filled with fragments it is inserted in data stream DS for transmission to the clients (step S120). The server 11 is arranged to continuously monitor if retransmission requests of missing fragments of an already transmitted data packet is received from a client (step S130). If a retransmission request REQ of missing fragments of a data packet is received (step S140), the missing fragments (here fragment [P(4):1]) are distributed into currently generated macro-packets MPc (step S150) which is packed with fragments associated with different subsequent fragmented packets originating from the data stream DS.sub.0, and then retransmitted in the new macro-packet MPc (step S120). The retransmission distribution of fragments may be arranged according to a preselected distribution scheme where the requested fragments are distributed e.g. in every n.sup.th macro-packet, or not retransmitted governed by time limitations. For instance, when the data stream is part of a live broadcast such as live TV distribution, retransmission of a missing fragment may be down prioritized if the expected arrival time at the client is too late for playout.

[0040] According to an embodiment of the invention, the generated macro-packets comprise subsequent fragments originating from one or more subsequent data packets of the data stream DS.sub.0 (or alternatively subsequent fragments of a continuous data stream). Thus, more than one fragments originating from different data packets may be transported in the same macro-packet, and the originating data packets need not be directly subsequent to each other (especially when retransmitting fragments). The packing order in the macro-packets is typically governed by a predetermined packing algorithm. In an exemplifying embodiment as illustrated in FIG. 7a, the method as described with reference to FIG. 4 will now be visualized starting with a data stream DS.sub.0 for transmission which comprises a number of subsequent fragments 1-X (each illustrated as a rectangle). At the sending side, subsequent fragments 1, 2, 3, 4, . . . 11, 12 of the data stream DS.sub.0 are packed into macro-packets MP1, MP2, MP3, each containing four fragments (N=4, S110). The macro-packets MP1, MP2, MP3 are transmitted as data stream DS which at the client side is received as data stream DS.sub.2 (S120). At reception of the data stream DS.sub.2, it is determined that the macro-packet MP2 is missing and the client device sends a request REQ for the missing packet MP2 (S130) to the sending side. Based on the request, the fragments 5, 6, 7, and 8 of the missing packet MP2 are then interleaved in the currently generated macro-packets MP4, MP5, MP6, and MP.sub.7 (of which MP6 and MP.sub.7 are not shown in FIG. 7a), respectively. MP.sub.5 is illustrated as being currently generated in FIG. 7a. A close up picture of MP.sub.5 in FIG. 7b reveals that the missing fragment 6 is added at the end of the macro-packet. The distribution scheme is here selected to add missing fragments at the end of a sequence of currently generated macro-packet MPH. The generated macro-packets are then transmitted in the data stream DS (S120).

[0041] Referring now to FIG. 5, according to an embodiment of the method, which has essentially the same steps 100-150 as described above with reference to FIG. 4, in an additional step (step 148), FEC fragments are provided for the requested missing fragments ([FEC:P(4):1]] in the example shown in FIG. 5) and packed into generated macro-packets. Typically, FEC is calculated for a group of fragments to be resent. One or more FEC fragments are generated and interleaved in different currently generated macro-packets in the same manner as the other fragments to be resent. The added FEC fragments for the missing fragments are thus typically sent in separate macro-packets such that if the originally missing fragment is lost once more in transmission, reconstruction of the missing fragments can be provided by means of the FEC fragment.

[0042] Traffic overhead may be defined as the ratio between retransmission network traffic incurred at the server by retransmission and error recovery and the normal network traffic incurred by delivering of the media data, e.g. video. If the recovery of lost packets induce 10% additional network traffic the traffic overhead is thus 0.1. The available network bandwidth is finite and the network is thus sensitive to a too big traffic overhead.

[0043] As previously mentioned, according to embodiments of the method, a step of providing selective retransmission of missing data packets or fragments is applied in the method. In FIG. 6, a step for providing selective retransmission SEL of fragments (step 146) is illustrated. The selective retransmission is according to an embodiment based on a priority ranking of data types to be retransmitted. The data types between which prioritization can be applied may be, with different prioritization ranking, at least one of audio data, complete video image data, and delta changes of images data. The reason for doing this is that the viewer is most sensitive to audio disturbances. Also, since the delta changes of pictures relates to the full picture, a lost full picture makes the delta changes useless. When the network capacity is decreasing, fragments corresponding to missing audio data may be selected to be retransmitted with normal retransmission rate, while fragments related to delta changes of images are down prioritized, optionally on the client device viewers own choice, not to compromise the transmission success rate of the audio of the video.

[0044] Video data is a continuous-media that must be presented at a predetermined playout rate meaning that data not arriving in time for being presented at the screen or display is useless. Therefore, the time frame for requesting and/or retransmission of missing packets or packets which have not yet arrived due to network delay is finite.

[0045] According to an embodiment of the method, the selective retransmission comprises at least one of performing retransmission if it is determined that missing fragments will be received within a predetermined time period, and performing retransmission with a rate control. In the first case the method may comprise a step of determining/predicting if the missing fragment is expected to reach the receiver in time for replay/playout or some other processing step, and then simply cancel retransmission of a fragment that is determined not to reach the client device in time for play out. The second option will retransmit missing data packets or fragments according to a preselected rate. According to an embodiment, rate control is provided by prioritizing which fragments are being retransmitted. If for example the round trip time is slightly higher the time for two retransmissions, the strategy is to prioritize retransmission of the fragments that are being retransmitted a second time over the ones being retransmitted a first time because they have less time than the ones being retransmitted the first time but can still make it in time. Segments being requested to be retransmitted a third time will not be retransmitted at all.

[0046] When and in which manner selective retransmission is to be applied in the method, is according to an embodiment of the method determined by monitoring characteristics of the network. Quality of Service, QoS, transmission capacity or delay in the network are suitable characteristics upon which the selective retransmission and/or priority ranking is activated. Preselected, or selectable, threshold values of the characteristics are applied to choose different retransmission strategies. The characteristics of the loss of data in a network might differ between different network types. For example, if the access to the network is based on fiber to the home the losses are typically sporadic and sharing of resources between different users are low meaning that in that case only severe losses shall impose aggressive selective retransmission. If the device is connected via mobile data network such as 3G, 4G or LTE, the characteristics of the losses and delays can depend on that the resources are shared with other users in the same mobile cell or that the device has entered an area with bad reception. This means that the retransmission strategy needs to be more careful not using the limited network resources for retransmitting data, thereby increasing the network load.