The present invention relates generally to communication networks and particularly to avoiding data loss in packet based networks.
In transmission of data packets over packet based networks, there is a possibility of packet loss, such that it may be assumed that a certain percentage of packets are lost on any packet based network. The percentage of lost packets is generally a function of the quality of the network. In some cases, such as transfer of a file, the loss of a small percentage of the transmitted data prevents the use of the entire file. In other cases, such as transmission of audio signals in accordance with specific formats, the loss of a small percentage of the transmitted data may go substantially unnoticed, or may cause only a small degradation in quality.
In many uses of packet based networks, an acknowledgment scheme is used, in which a receiver notifies a transmitter, which packets were received or were not received and therefore require retransmission. In some cases, however, use of an acknowledgment scheme is not desired. For example, in providing real time audio or video, allowing for sufficient time to request retransmission adds substantial delay to the transmission. In addition, acknowledgment schemes add substantial complexity to both the receiver and the transmitter. Acknowledgment schemes are especially problematic for multicast transmissions, where a single packet is transmitted to a plurality of receivers. The number of retransmission requests for a multicast packet stream on a packet based network may flood the network.
Instead of using an acknowledgment scheme, or in addition to an acknowledgment scheme, transmission of redundant data packets may be used to reduce the chances of losing data. For example, each packet may be transmitted twice. Such redundant transmission, however is relatively wasteful in bandwidth. In order to limit the amount of bandwidth used for the redundant transmission, a forward error correction (FEC) method may be used. The use of a FEC method allows a higher chance of being able to reconstruct the lost data, with a lower percentage of extra transmitted data. The use of a FEC method, however, requires large processing resources for decoding. FEC methods may be used in one or more transmission layers. For example, A FEC method may be used in the network layer and/or in the application layer.
In the enhanced data GSM evolution (EDGE) cellular system, different FEC codes are used for the headers of the transmitted packets and for the data of the packets. The FEC of the header is optionally stronger than the PEC of the data, so as to reduce the chances of loss of the header beyond the chances of loss of data. Thus, even when a packet is lost, the header may be salvaged, giving some information on the data of the lost packet.
An aspect of some embodiments of the invention relates to segmenting, for transmission, a stream of data into packets in a manner which takes into account an identification of data portions which are important for reassembly of the data stream in one or more higher layers. Optionally, the loss of one or more of the important data portions may cause an upper layer in a receiver to drop one or more data portions that were received by the receiver. In some embodiments of the invention, the important data portions comprise headers and/or tails of higher layer packets.
In some embodiments of the invention, the segmentation is performed in a manner which minimizes the number of segments including important data portions (these segments are referred to herein as important segments). Optionally, important data portions, such as headers, are fitted into a single segment when possible. Alternatively or additionally, important data portions are included in a plurality of segments.
Optionally, important data portions are not positioned at the end of a data segment and/or at the beginning of a data segment. In some embodiments of the invention, the data is loaded into segments, which indicate the end of packets of a layer directly above the segmenting layer. Optionally, the segmentation is performed such that the upper layer packet does not end at the end of the segment, so that the beginning of the next upper layer packet is located in the same segment as the indication on the position of the beginning of the packet. Thus, the indication may be protected from loss, as described below, together with the important data, in a single segment.
Optionally, the identification of the important segments is performed by receiving information on the structure of the data from upper layer units. This involves interrelation between layers which, for simplicity, is generally avoided in prior art communication systems.
An aspect of some embodiments of the invention relates to a method of transmitting segmented data over a lossy network. Segments including data portions identified as being important for reassembly (these segments are referred to herein as important segments) are optionally transmitted in a manner which reduces their chance of being lost.
In some embodiments of the invention, important segments are transmitted on a channel with a lower error rate relative to the remaining data. For example, important segments may be transmitted on a channel using a first modulation method, while other segments are transmitted on a channel using a second modulation method with a higher error rate, for example BPSK In an exemplary embodiment of the invention, important segments are transmitted on a wire link, while other segments are optionally transmitted on a parallel wireless link. In another exemplary embodiment of the invention, a plurality of channels (for example wireless channels) are used for transmitting the data. Optionally, the quality of the channels is periodically determined and according to the determination a high quality channel is selected for important segments.
Alternatively or additionally, the important segments are transmitted with redundancy. Optionally, important segments are transmitted in a plurality of copies. The plurality of copies of the important segments are optionally substantially identical. Alternatively or additionally, one or more of the copies includes substantially only the important data of the segment.
Alternatively or additionally, important segments are transmitted with a forward error correction (FEC) scheme or with a stronger FEC than used for non-important segments.
In some embodiments of the invention, the redundancy transmission is performed by a transmitter without the receiver being aware of the redundant transmission. The transmission is optionally performed such that the receiver will discard the redundant data as junk if it is not required. Optionally, the communication between the transmitter and receiver is governed by a protocol (or an operation mode of the protocol) which does not support redundant transmission. The transmitter prepares the redundant transmissions in a manner which will allow them to be used by the receiver when necessary even though the receiver was not configured to receive redundant transmissions.
In some embodiments of the invention, the transmission of data in the layer in which the segmentation is performed is unreliable, i.e., there is no guarantee that all the transmitted data was actually received by the receiver. Optionally, the receiver does not transmit acknowledge messages at all. Alternatively, the receiver transmits some acknowledge messages, for example for use as feedback in channel adaptation. In some embodiments of the invention, the entire transmission of the data in all the layers is unreliable. Alternatively, one or more layers above the segmenting layer, are reliable.
An aspect of some embodiments of the invention relates to a method of generating a forward error correction (FEC) code for transmitted data, which method takes into account the lower layer segmentation of the data.
In some embodiments of the invention, the FEC code comprises an erasure FEC code which takes into account the segmentation of the packets and hence is aware of which portions of data will be lost together.
In some embodiments of the invention, the FEC is generated while taking into account an uneven segmentation for transmission, of the data. In some embodiments of the invention, the FEC is generated on a version of the data which includes padding bits in positions selected according to the segmentation of the data for transmission. Optionally, the padding bits are inserted in place of headers added to the data for the transmission, such that each FEC unit is transmitted in a separate transmission segment which may be lost, e.g., RLC segment. Optionally, some of the data units used in calculating the FEC are shorter than others (they are padded with filler bits), while the FEC algorithm operates on equal size data units. The receiving unit optionally is configured with the segmentation of the data, for using the FEC.
An aspect of some embodiments of the invention relates to transmitting information in a cellular network to one or more mobile stations. The information is represented by a plurality of different data sequences, optionally including redundancy, such that at least two of the cells of the network transmit different data sequences.
In some embodiments of the invention, the information is represented by a plurality of data units, such that the information can be reconstructed from any combination of received data units, including at least a specific number of non-correlated data units. Optionally, different cells of the network, particularly adjacent cells, transmit different combinations of units. Thus, a mobile station moving between cells, possibly loosing some of the data units due to the movement, has a higher chance to collect a sufficient number of different data units. The transmission of different data units reduces the chances of receiving the same data units in two different cells due to a timing difference (e.g., loss of synchronization) between the cells.
Alternatively or additionally, measures are taken to prevent loss of synchronization between cells. In some embodiments of the invention, the cells are required to use the same bandwidth allocation in transmitting same data. Alternatively or additionally, a cell having a lower bandwidth allocation for the transmitted data is required to discard packets in order to keep synchronization with the other cells. The discarded packets are optionally redundant packets from each block of packets, such that data is not necessarily lost, although the chances of decoding the data in case additional packets are lost, is reduced.
An aspect of some embodiments of the invention relates to a method transmitting control information between a mediator and a receiver, in-band, in a cellular network. The control information is transmitted in specific segments which include important data. Optionally, measures are taken to reduce the chances the important data will be lost and therefore the chances that the control information will be lost is reduced. Alternatively or additionally, if the important data is lost the control information is not required.
There is therefore provided in accordance with an exemplary embodiment of the invention, a method of transmitting data, comprising receiving data to be transmitted to a receiver, determining information on the structure of the received data, segmenting the received data into segments, in a manner determined responsive to the determined structure information; and transmitting the segments to the receiver.
Optionally, determining information on the structure of the received data comprises receiving the structure information from an upper communication layer providing the data. Optionally, determining information on the structure of the received data comprises determining the structure information from the received data. Optionally, determining the structure information from the received data comprises examining the received data for predetermined bit sequences. Optionally, determining information on the structure of the received data comprises identifying portions of the data which if not received by the receiver will have an adverse affect on utilizing at least one portion of the data received by the receiver.
Optionally, identifying portions of the data comprises identifying packet headers of protocol layers above the segmenting layer. Optionally, identifying packet headers comprises receiving data from an upper communication layer and/or scanning the data for sequences of headers. Optionally, segmenting the data into segments comprises segmenting in a manner which minimizes the number of segments including identified portions. Optionally, segmenting the data into segments comprises segmenting such that identified portions are included in a single transmitted segment. Optionally, the method includes padding with filler bits, one or more upper layer packets of the received data, if necessary, in order to ensure that identified portions are included in a minimal number of transmitted segments.
Optionally, segmenting the data into segments comprises not positioning identified portions in the beginning of segments. Optionally, identifying portions of the data comprises identifying border portions of protocol layers above the segmenting layer, including a tail of a previous packet and a header of a following packet. Optionally, segmenting the data into segments comprises segmenting such that border portions are included in a single segment and/or identified border portions appear at the beginning of segments. Optionally, segmenting the data into segments comprises segmenting such that at least some of the segments include at least one complete upper layer packet.
Optionally, segmenting the data into segments comprises segmenting such that at least some of the segments includes a single upper layer packet. Optionally, segmenting the data into segments comprises including at least one specific identified portion in more than one segment.
Optionally, transmitting the segments comprises transmitting at least some of the segments including identified portions in a manner which reduces the chances that their data will be lost, relative to at least some other segments. Optionally, transmitting the segments comprises transmitting segments including identified portions separated by at least a predetermined number of segments not including identified portions.
There is further provided in accordance with an exemplary embodiment of the invention, a method of transmitting data, comprising receiving data to be transmitted to a receiver, segmenting the received data into segments, identifying segments including received data which is important for reassembly, and transmitting segments identified as including important data, with a reduced loss rate, relative to at least some of the non-identified segments.
Optionally, identifying segments including received data which is important for reassembly comprises identifying portions of the data which if not received by the receiver will have an adverse affect on utilizing data in at least one segment received by the receiver. Optionally, identifying segments including received data which is important for reassembly comprises identifying headers of one or more upper layer protocols. Optionally, identifying segments including received data which is important for reassembly comprises identifying based on information from an upper layer communication layer providing the data. Optionally, identifying segments including received data which is important for reassembly comprises identifying based on an examination of the received data. Optionally, transmitting segments identified as including important data, with a reduced loss rate comprises transmitting segments including important data on a low error rate channel. Optionally, transmitting segments identified as including important data, with a reduced loss rate comprises transmitting for segments including important data, at least one additional segment including the important data of the segment. Optionally, transmitting the at least one additional segment comprises transmitting an additional segment including substantially the same data as the original segment. Optionally, transmitting the at least one additional segment comprises transmitting an additional segment including substantially only the important data of the original segment. Optionally, transmitting the at least one additional segment comprises transmitting at least two segments including the important data of the original segment, each of the at least two segments including a portion of the data of the original segment not included by the others of the at least two segments.
Optionally, transmitting the at least one additional segment comprises transmitting the at least one additional segment separated from the transmission of the original segment by at least a predetermined number of segments or at least a predetermined amount of time. Optionally, the receiver is not adapted to handle the transmitted at least one additional segment. Optionally, transmitting segments identified as including important data, with a reduced loss rate comprises transmitting segments including important data with a forward error correction code.
There is further provided in accordance with an exemplary embodiment of the invention, a method of transmitting a data block to one or more cellular stations, comprising generating a plurality of different data sequences representing the data block and transmitting each of the different data sequences in a different cell of a cellular network servicing the cellular stations.
Optionally, generating the plurality of different data sequences comprises encoding the data block into a plurality of different data sequences. Optionally, generating the plurality of different data sequences comprises generating data sequences including the data block together with one or more forward error correction FEC units. Optionally, generating the plurality of different data sequences comprises generating a master sequence, and generating the plurality of different data sequences by forming sub-groups of the master sequence. Optionally, generating the master sequence comprises generating a first number of data units, such that the data block can be reconstructed from any sub-group of the master sequence including at least a second number of data units. Optionally, each transmitted data sequence includes at least a third number of data units, greater than the second number. Optionally, transmitting each of the different data sequences in a different cell comprises transmitting responsive to instructions generated by a central unit at substantially the same time. Optionally, transmitting each of the different data sequences in a different cell comprises transmitting the different data sequences such that their transmission times at least partially overlap.
Particular non-limiting embodiments of the invention will be described with reference to the following description of embodiments in conjunction with the figures. Identical structures, elements or parts which appear in more than one figure are preferably labeled with a same or similar number in all the figures in which they appear, in which:
It is noted that the network of
Data transmitted from Internet 60 to an MS 20 is optionally transmitted to GGSN 58, from the Internet, in IP packets. GGSN 58 transfers the IP packets to SGSN 52 which optionally parses the IP packets into logical link control (LLC) packets, as described below with reference to
In some embodiments of the invention, the receiving MS 20 does not acknowledge the receipt of the RLC segments and does not request retransmission of RLC segments not received. This type of non-acknowledgment operation mode may be advantageous, for example, when -real time data is transmitted to an MS 20 and/or when multicast data is transmitted to a plurality of MSs 20. Alternatively or additionally, a non-acknowledgment operation mode may be used when an application layer handles retransmission and/or uses a FEC method.
Alternatively to transferring the data through GGSN 58, SGSN 52 and/or BSC 34, and optionally changing one or more thereof to implement the present invention, a bypass unit (referred to herein as a point to multi-point unit (PTMU) 49 located between BSC 34 and BTSs 32 emulates the operation of GGSN 58, SGSN 52 and/or BSC 34 to provide data in accordance with the present invention. Optionally, a mediator 53 receives data from a content provider, directly or through the Internet. The mediator optionally partially prepares the data for transmission and provides the data to PTMUs 49 of the cells in which the data is to be transmitted. The PTMUs 49 optionally complete the processing and provide the resultant RLC segments to BTSs 32 for transmission to MSs 20. An exemplary division of the tasks between mediator 53 and PTMUs 49 is described hereinbelow.
LLC packets 220 are transmitted to BSC 34 where they are optionally filled into RLC segments 225 of an RLC segment stream 222, transmitted to MS 20. Each RLC segment 225 is optionally formed of an RLC data portion 226 and an RLC header 228. Header 228 optionally includes a block sequence number (BSN) of, for example, 7 bits which indicates the position of the RLC segment 225 within the stream 222. In addition, header 228 indicates the position at which an LLC packet 220 ends within the RLC segment 225, if such an LLC end exists in the RLC segment.
Alternatively to the tasks of
As is known in the art, the IP packets 210 are parsed (306) into SNDCP packets which in turn are converted (308) into LLC packets 220. In some embodiments of the invention, the parsing (306) of IP packets into SNDCP packets is performed in a manner which allows simple secure transmission of the IP header and/or the UDP header of the IP packet. Optionally, the parsing (306) is performed such that the IP and/or UDP header are included in one or more SNDCP packets separate from one or more other SNDCP packets which carry the data of the packet. Thus, it is easier to apply high quality transmission to header portions of the IP packets.
In some embodiments of the invention, the headers of the IP packet are parsed into SNDCP packets of a small size, such that each resultant LLC packet fits into a single RLC segment. The RLC header of the segment optionally identifies the beginning and the end of the LLC packet. In an exemplary embodiment of the invention, each RLC segment including IP header data includes an RLC header followed by an LLC header, an SNDCP header, IP header data and an LLC tail. Alternatively, all the IP header data of an IP packet is included in a single LLC packet.
The data of the IP packet is optionally parsed (306) into SNDCP packets of maximal size, in order to minimize the overhead involved in the SNDCP header and the LLC packet header and tail. Alternatively or additionally, when padding (304) of IP packets is required for proper fitting into the RLC segments, when possible, the padding is replaced by a breakup of one or more SNDCP packets into a plurality of packets.
The LLC packets 220 are then segmented (310) into RLC segments 225, as is known in the art. As mentioned above, an RLC header 228 is generated for each RLC segment 225. The RLC header 228 identifies the end points of LLC packets included in the RLC segment. In some embodiments of the invention, if (312) an RLC segment ends with a portion of an LLC border portion, without including the entire border portion, the LLC packet is optionally enlarged in order to avoid the breaking of the LLC border portion between two RLC segments 225. The LLC packet is optionally enlarged by adding padding bits at the end of the packet. The padding bits are optionally of a predetermined form which may be easily identified and removed by the application layer of the receiver. Optionally, messages including inadvertently the padding bits are slightly altered so that they will not be interpreted as padding bits. Alternatively, the LLC packet is enlarged by transferring data bytes from a following LLC packet to the current LLC packet. Alternatively, as described below, the sizes of the LLC packets and the IP packets are predetermined, so that the LLC border portion is always included in a single RLC segment.
In some embodiments of the invention, LLC packets are enlarged such that LLC border portions always appear at the beginning of an RLC segment. Having the LLC border always appear at the beginning of an RLC segment may simplify the operation of the application layer in MS 20, for example may simplify the calculations required for a forward error correction (FEC) code of the data. The FEC calculation is optionally simplified by having only two possible application data layouts in RLC segments. In a first layout, the entire data portion of the RLC segment includes application data and in a second layout the entire RLC data except for a first few bytes used for the LLC border portion are used for application layer data. Alternatively to enlarging the LLC packet in order to avoid transmitting an LLC border in two RLC segments, redundant transmission is performed for both the RLC segments.
In some embodiments of the invention, each generated RLC segment 225 is optionally checked (314) to determine whether the packet includes an LLC border, an IP header, a UDP header and/or portions thereof. The headers are optionally considered important portions, which should be transmitted to the receiver with a lower chance of loss, as their loss may cause inability to reconstruct large amounts of received data beyond the lost data.
RLC segments determined to include an important portion are optionally transmitted with redundancy (316). In some embodiments of the invention, for RLC segments 225 determined to include an important portion, an additional redundant RLC segment, including a copy of the important portion, is optionally generated, so as to reduce the chances of loss of the important portion. The LLC borders generally include sufficient information to allow reconstruction of an IP packet, even if some of the data of the packet was lost. Thus, the affect of the loss of a non-important RLC segment does not go beyond the actually lost bits. It is noted that in many cases, the loss of an LLC border would cause the loss of the entire IP packet in which the data of the lost LLC border was included.
The generated RLC segments (original and additional) are optionally transmitted (318) to MS 20, which regenerates IP packets from the RLC segments it receives. In some embodiments of the invention, the original and additional copies of an RLC segment are transmitted one after the other in order to allow simple association of the copies. Alternatively, for example when the receiver has a reception window of a sufficient size, the copies of a same RLC segment are transmitted with sufficient separation, such that the chances of loss of all the copies in a single burst of noise is relatively small. The separation is optionally of at least a predetermined amount of time and/or a predetermined number of intervening transmitted RLC segments. Alternatively or additionally, different RLC segments including important portions are transmitted, when possible, separated by a few non-important RLC segments, such that in case of damage by a bursty noise, the number of lost segments including important portions is minimal.
As to determining (302) where the IP packet will end in the RLC segment stream 222, the determination is optionally performed based on the known sizes of the LLC and RLC segments to be generated. In some embodiments of the invention, as described below, a predetermined layout of LLC packets and RLC segments is used, and the application data is fit into predetermined spaces of the predetermined layout, that are assigned for application layer data.
Referring in more detail to padding (304) the IP packet with filler bits, the padding is optionally performed such that the beginning of each IP packet is aligned with the beginning of a new RLC segment 225, for reasons discussed below. Inserting the IP header from the beginning of an RLC segment, may reduce the number of RLC segments 225 required for the IP header and hence requiring additional transmission for redundancy. Thus, aligning the beginning of the IP packet with the beginning of an RLC segment reduces the bandwidth required for redundancy.
The padding is optionally performed by adding a required number of bits after the IP packet, without changing the length indication of the IP packet. Thus, when received by MS 20, it will discard the padding bits as not belonging to any IP packet.
Alternatively or additionally to padding the IP packet, the size of the LLC packets are changed dynamically according to the sizes of the transmitted IP packets, such that the IP headers are always placed in RLC segments in a desired manner. Further alternatively or additionally, IP headers 206 are allowed to be included in RLC segments with non header data of the same IP packet and/or of a different IP packet. Optionally, redundant transmission is performed for any RLC segment including even a part of an IP header.
Referring in more detail to identifying packets including headers (314,
In some embodiments of the invention, the additional RLC segments are exact copies of the original RLC segments for which they were generated, including the same block sequence number (BSN). If one of the original and additional packets is lost on its way to MS 20, the deciphering of the packet will proceed without problems. If, however, both the packets are received consecutively, MS 20 will assume that the second packet to be received is a different packet from the first to be received, as is now described.
During reconstruction, a first LLC packet 430 includes the data of RLC segment 404 and optionally data from previous RLC segments and the first portion 415 of RLC segment 401. The data included in LLC packet 430 ends at the end of portion 415, as indicated by pointer 410 of RLC segment 401. A following LLC packet 432 includes the second portion 418 of RLC segment 401 together with the first portion 425 of RLC segment 402. The data included in LLC packet 432 ends at the end of portion 425, as indicated by pointer 410 of RLC segment 402. Since MS 20 assumes that packets 401 and 402 are different, although they have the same BSN, MS 20 assumes that a series of RLC segments were lost. Therefore, MS 20 optionally adds to LLC packet 432, in a field 435, padding bits to replace the assumed missing RLC packets. The second portion 428 of RLC segment 402 is included in a further LLC packet 434, together with data from one or more subsequent RLC segments.
In checking LLC packet 432, a CRC check of the header of the packet will fail, as the tail including the CRC belongs to a different packet than the LLC header. Therefore, LLC packet 432 will be discarded. Thus, the data provided to an upper layer includes first portion 415 and second portion 428, covering exactly one copy of the transmitted data.
In some embodiments of the invention, the transmission times of the RLC segments are adjusted, in order to minimize the chances that the additional packet will be received by MS 20 after one or more subsequent packets are received. Optionally, the original packet and its duplicate are transmitted one after the other without delay and thereafter, a short, but substantial, delay is added before transmitting the subsequent RLC segment. Alternatively or additionally, the packets are transmitted on a communication channel which does not allow a different order of packets in reception than in transmission. For example, the RLC segments may be transmitted in time slots, and the time slots used for the RLC segments are ones which have fixed relative timings.
When both the original and additional copy of an RLC segment including a portion of an IP header are received by MS 20, each RLC segment will be converted into a separate LLC packet (as described above, the IP header is optionally broken up into separate LLC packets corresponding to respective RLC segments). The duplicate LLC packets will then optionally be identified by the LLC layer. In those embodiments in which the entire IP header is included in a single LLC packet, the duplicate RLC segments for all the RLC segments carrying data of the header are optionally transmitted after all the original RLC segments. If all the segments are received the duplication will be identified by the LLC layer. If one of the sets of RLC segments is received, even if one or more of the duplicates is lost, the IP header will be received by the LLC layer and the duplicates will be dropped.
Alternatively to the additional RLC segment having the same BSN as the original RLC segment it copies, the additional packet carries a consecutive BSN value to the value of the original copied RLC segment, as is now described.
In reconstructing LLC packets, a first LLC packet 530 includes the data of RLC segment 504 (and optionally data from previous RLC segments) and the first portion 415 of RLC segment 501. The data included in LLC packet 530 ends at the end of portion 415, as indicated by pointer 410 of RLC segment 501. A following LLC packet 532 includes the second portion 418 of RLC segment 501 together with the first portion 425 of RLC segment 502. The data included in LLC packet 532 ends at the end of portion 425, as indicated by pointer 410 of RLC segment 502. Unlike the embodiment of
In checking LLC packet 532, the CRC check fails as the tail does not match the header and therefore LLC packet 532 will be discarded. Thus, the data provided to an upper layer includes first portion 415 and second portion 428, covering exactly one copy of the transmitted data. Alternatively or additionally, proprietary measures are taken to make sure that the CRC check of packet 532 will fail.
If one of segments. 501 or 502 is lost, MS 20 will add padding bits instead of the “missing” RLC segment. The application layer optionally identifies the padding bits and removes them. In some embodiments of the invention, the application layer on MS 20 is aware of the segmentation structure and accordingly identifies the padding bits, as data is lost in entire RLC segments. The segmentation structure may be predetermined and/or may be transmitted to the receiver with each IP packet, group of IP packets and/or periodically, for example within the IP headers.
In some embodiments of the invention, the identification of the padding bits is performed by the application layer on MS 20 after the data is reconstructed into IP packets. In such cases, when padding bits are added, the end of the IP packet will be truncated according to the size of the IP packet indicated in the IP header. Optionally, in order to avoid loss of data, a predetermined number of bytes at the end of each IP packet including an important segment are padded with meaningless bytes. The number of meaningless bytes is optionally selected according to the number of important segments in the IP packet and/or the average loss rate on the transmission link to MS 20. Optionally, the meaningless padding bytes are not transmitted and instead the RLC numbers of the meaningless padding bytes are skipped. Thus, MS 20 will think that one or more RLC segments were lost and will replace them with padding bits. If additional padding bits were added in the middle of the IP packet, some or all of the padding bits at the end will be discarded due to the length of the IP packet. Otherwise, the padding bits at the end of the IP packet will be removed by the application layer.
Alternatively or additionally, an identification marker, for example having a length of 2-4 bytes, is added to each RLC segment. The application layer can then determine which RLC segments were lost according to the identification markers of the received RLC segments. This alternative is especially advantageous in those embodiments in which RLC segments of different lengths are used, and the receiver does not always know the number of bytes lost with an RLC segment.
As mentioned above with reference to
Alternatively to always giving the same BSN to the copy segment as to the original segment or always giving a different BSN, at different time points and/or for different packets, different policies for assigning BSN numbers to copy segments, are used. For example, copies of RLC segments including IP header portions may be assigned a different BSN than the original segment, while copies of RLC segments including LLC border portions may be assigned a same BSN as the original segment.
Alternatively to the additional RLC segment (e.g., 402, 502) including all the bits of the original segment (e.g., 401, 501), the additional segment includes only some of the bits of the original segment (e.g., the important bits), and the remaining portion of the RLC segment is used for additional data, thus reducing the total number of transmitted RLC segments. In some embodiments of the invention, the additional RLC segment includes only the bits which caused the additional packet to be generated, for example, LLC header bits and/or IP header bits.
In some embodiments of the invention, in preparing the RLC segments for transmission, the sizes of the LLC packets are selected such that the border portions fit at the end of a first RLC segment and are repeated at the beginning of a subsequent RLC segment. Alternatively to repeating the entire border portion, only the LLC tail or only the LLC header are repeated. This alternative reduces the bandwidth required for redundancy but may cause additional loss of data, in case one of the RLC segments is lost.
In order to reduce the damage in case an important RLC segment (the following description relates to segment 642 but is true for substantially any other important RLC segment) was lost (even though it was transmitted with redundancy), in some embodiments of the invention, all the regular LLC packets 644 and 646 have the same checksum value (referred to in
In some embodiments of the invention, in order to prevent the receiver from considering adjacent LLC packets 644 and 646 having the same checksum value as duplicate copies of a same LLC packet, each RLC segment including an LLC border portion 634 includes an additional short LLC packet 639 with a different checksum value (referred to in
If RLC segment 642 is lost, the short LLC packet 639 is lost and the two LLC packets 644 and 646 of the border portion are considered as a single LLC packet. If the important RLC segment 642 is not lost, the short LLC packet 639 is received between the two LLC packets 644 and 646 of the border portion, such that the two LLC packets 644 and 646 of the border portion having the same checksum are not received one after the other and are not considered redundant.
In some embodiments of the invention, the short LLC packet 639 includes a minimal amount of data 657, e.g., one byte, required so that the short LLC packet is considered a valid packet. Alternatively, the short LLC packet includes a larger amount of data, for example filling the remaining portion of the RLC segment. Optionally, the short LLC packet includes data of the IP packet being transmitted. Alternatively, the short LLC packet includes meaningless filler bytes, removed by the application layer. Alternatively to the filler bytes being removed by the application layer, the short LLC packets 639 include an indication of a non-existent IP layer, which will cause the short LLC packets 639 to be discarded at the receiver.
It is noted that although LLC tail 643 is shown immediately after RLC header 228, this is not required, and data may be included in important RLC segments 632, 642 and/or 652, between RLC header 228 and border portion 634. Similarly, although important RLC segments 632, 642 and 652 are shown as including data after LLC header 645, this is not necessary, and the segments may end with LLC headers 645.
In some embodiments of the invention, additional packets for redundancy are transmitted for only some of the important packets, in order to reduce the bandwidth used for redundancy. Optionally, the headers for which additional copies are generated are selected randomly so as to provide an additional copy with a predetermined probability. Alternatively or additionally, the additional packets are generated for headers of relatively large packets and/or when the amount of data to be lost due to the loss of the header is relatively large.
Alternatively or additionally to transmitting redundant segments for important segments, in some embodiments of the invention, important segments are transmitted with a stronger coding scheme. For example, a mode in which RLC segments of different data lengths (the remaining portions of the RLC being used for redundancy coding) are used for redundancy transmission in accordance with the present invention. For example, regular data may be transmitted in RLC segments including 36 or 30 data bytes, while important segments are transmitted in RLC segments including 20 or 24 bytes of data. In this alternative, when data is lost, the application layer in the receiver optionally determines the amount of data lost according to the location of the lost RLC segment in the transmitted sequence, based on a predetermined arrangement of the transmissions.
It is noted that in the above description MS 20 is not adapted to handle redundant transmissions. Thus, the invention may be implemented without changing MSs 20. In some embodiments of the invention, however, MSs 20 are configured to identify redundant packets. In these embodiments, the operation of MSs 20 is simplified and they do not need to form LLC packets which are later discarded. Optionally, the transmission of packets is performed in a manner which is compatible with MSs 20 which are not adapted to handle redundant transmissions, and those MSs 20 which are adapted to identify the redundancy enjoy the advantages of simpler operation. In some embodiments of the invention, the MSs 20 identify the redundant packets by comparing the packets received to see that they are identical. Alternatively or additionally, any other method which does not include adding markings to the packets is used by MSs 20 to identify the redundant RLC segments. Further alternatively or additionally, markings are added to the redundant RLC segments, and/or to the packets for which copies were prepared, so as to allow easy identification by MSs 20 configured to identify the packets.
In some embodiments of the invention, the markings require that all the MSs 20 are configured to identify the redundant packets. These embodiments allow simpler methods of redundancy, but require configuration of all MSs 20 used.
Instead of generating redundant segments and/or in addition to the redundant segments, important RLC segments may be transmitted on a channel having a higher quality than the channel used for regular RLC segments. Optionally, the connection uses two channels for transmission of data. One of the channels has a higher bit rate and therefore a lower quality. Optionally, important segments are transmitted on the higher quality channel. Optionally, when there is additional bandwidth on the high quality channel, regular segments are transmitted on the high quality channel. Alternatively, regular segments are not transmitted on the high quality channel in order to further increase the quality level of the channel.
In some embodiments of the invention, as mentioned above, the application data transmitted to MS 20 is protected by a forward error correction (FEC) code, such that if some of the transmitted RLC segments are lost the data can still be reconstructed from those RLC segments which were received The FEC is optionally prepared with the data of each RLC segment serving is a separate unit of the FEC preparation. Since some RLC segments include an LLC border portion, these segments include less data. Optionally, in order to have equal size data units for FEC calculation, the data of which is to be packaged into a segment including an LLC border portion is padded with padding bits for the purpose of calculating the FEC. The padding bits are not transmitted to MSs 20.
Each IP packet 588 is optionally loaded into a predetermined set 592 of RLC segments 225. In an exemplary embodiment of the invention, each set 592 includes a predetermined number of RLC segments 225 (marked 225A, 225B and 225C) which are pre-allocated to specific payloads. For example, two first RLC segments 225A carry header 590, while the remaining RLC segments 225 carry the blocks 582 and 586. Optionally, some of these RLC segments (marked 225B) include LLC border portions 634 (
In some embodiments of the invention, each RLC segment carries a single data unit, which may either be a FEC data unit or an original-data unit. RLC segments 225B including LLC border portions 634 (
In an exemplary embodiment of the invention, each data block 582 together with its respective FEC block 586 is fit into 15 IP packets, wherein each IP packet is fit into 17 RLC packets.
The use of a predetermined segmentation of the data and the loss of data only in predetermined patterns, allows the use of a simpler FEC, such as an erasure FEC. Such a FEC method does not require the determination of where the error occurred and therefore requires much less processing resources than if the finding of errors was required. In addition, such a FEC method requires lower redundancy rates of data transmission for achieving same chances of successful decoding of the data.
When data is received (553) by mediator 53 for transmission, the original data 580 is fitted (554) into blocks of the predetermined size. In some embodiments of the invention, if necessary, the original data 580 is padded (556) with padding bytes 584 to include an integer number of original data blocks 582. Alternatively, the remaining data is handled in accordance with embodiments described above not involving predetermined segmentation. Further alternatively or additionally, predetermined segmentations are prepared for a plurality of block sizes and the block size used is selected according to the amount of data transmitted.
Thereafter, FEC units are prepared (558) for the data. The FEC is optionally prepared using any method known in the art. Optionally, in preparing the FEC units, original-data units shorter than others (e.g., corresponding to segments 225B including LLC border portions) are padded with padding bits (not shown in
Original data 580 and FEC units 586 are optionally filled (560) into IP packets 588 in an appropriate order, such that each data unit will fit into its pre-planned RLC segment 225. In some embodiments of the invention, the filled IP packets 588 are transferred (562) to PTMUs 49 of the cells in which they are to be transmitted. Each PTMU 49 optionally prepares the IP packets for transmission, as described above with reference to
In some embodiments of the invention, the data is filled (560) into IP packets 588 according to the order in which they are generated, such that they will be transmitted in the order they were generated. Alternatively, the data of different blocks 582 is interleaved, so that in case a burst of consecutively transmitted segments are lost, not all the lost segments are from a single block. MSs 20 are optionally configured with the exact division used, for example as shown in
In some embodiments of the invention, one or more of the determined parameters of the FEC are transmitted to the receiver MSs 20 in one or more of the RLC segments transmitted with redundancy 225A or 225B. MSs 20 are optionally configured to identify these parameters. The positions in which these parameters are encoded are optionally excluded from participating s in the calculations of the FEC (these positions do not contain original data or FEC data). In some embodiments of the invention, the parameters are provided in segments 225A which anyhow do not carry application data. As described above, segments 225A are transmitted with redundancy and therefore their chances of being lost are small. Furthermore, in some cases, when segments 225A or 225B are lost the entire IP packet cannot be deciphered and therefore the parameters are not needed.
In some embodiments of the invention, the same original data is to be transmitted in a plurality of different cells. Optionally, in order to avoid an MS 20 moving between cells to receive the same FEC units twice, which in case of data loss may prevent reconstruction of the original data, different cells transmit different sequences representing the same original data. Optionally, for each block of the original data, mediator 53 generates a larger number of non-correlated FEC units than required for FEC block 586 of the original data block. The original data together with all the generated FEC units are referred to together as a master sequence. For each cell, mediator 53 optionally generates IP packets using the original data and a sub-group of the FEC units in the master sequence. In some embodiments of the invention, the FEC units of the cells are selected such that adjacent cells have as few as common FEC units as possible.
Alternatively to all the cells receiving all the original data, some of the cells may receive less than all the original data and instead receive a higher proportion of FEC units. Thus, the difference between the content transmitted in adjacent cells may be enlarged. It is noted that different cells may receive the same ratio of FEC units versus original data units or different ratios.
In some embodiments of the invention, instead of transmitting different FEC units in different cells or in addition, the cells are instructed to transmit the data they receive in synchronization. Optionally, a cell which is not able to transmit the data at a predetermined rate discards some of the extra FEC units and/or original units it receives from each block, in order to be synchronized with other cells.
The above exemplary embodiment involves a cellular network, which generally has high packet loss rates, e.g., about 10%. It is noted, however, that the present invention is not limited for use with any specific network type or for any loss rate, although its importance increases with the loss rate of the network.
The predetermined time between transmitting the original packet and transmitting the image packet 606 is optionally chosen to be relatively long such that the chances of image packet 606 being received before the original packet are as slight as possible, but not too long, so that if the original packet is lost, the image packet 606 is received before a time-out timer expires. Optionally, the predetermined time between transmitting the original packet and transmitting the image packet 606 is chosen based on measurements of the round trip delay of the connection.
If both the original packet and the image packet are received, the image packet being received later, the image packet will be discarded as being redundant. If only the image packet is received, the image packet may be used by the receiver to determine how much data was lost and/or the type of data lost.
In some embodiments of the invention, as described above, an image packet is generated for each packet to be transmitted. Alternatively, image packets 606 are generated only for some of the packets, for example for relatively long packets and/or for packets having a higher importance. Alternatively or additionally, image packets 606 are generated for a certain percent of randomly selected packets.
It will be appreciated that the above described methods may be varied in many ways, including, changing the order of steps, and the exact implementation used. The methods of the present invention may be performed in various protocol layers and may be performed for a single transmission system in a plurality of communication protocol layers. It should also be appreciated that the above described description of methods and apparatus are to be interpreted as including apparatus for carrying out the methods and methods of using the apparatus.
The present invention has been described using non-limiting detailed descriptions of embodiments thereof that are provided by way of example and are not intended to limit the scope of the invention. It should be understood that features and/or steps described with respect to one embodiment may be used with other embodiments and that not all embodiments of the invention have all of the features and/or steps shown in a particular figure or described with respect to one of the embodiments. Variations of embodiments described will occur to persons of the art.
It is noted that some of the above described embodiments may describe the best mode contemplated by the inventors and therefore may include structure, acts or details of structures and acts that may not be essential to the invention and which are described as examples. Structure and acts described herein are replaceable by equivalents which perform the same function, even if the structure or acts are different, as known in the art. Therefore, the scope of the invention is limited only by the elements and limitations as used in the claims. When used in the following claims, the terms “comprise”, “include”, “have” and their conjugates mean “including but not limited to”.
Number | Date | Country | Kind |
---|---|---|---|
154739 | Apr 2003 | IL | national |
Filing Document | Filing Date | Country | Kind | 371c Date |
---|---|---|---|---|
PCT/IL04/00204 | 3/3/2004 | WO | 10/24/2006 |