WO2014043908A1

WO2014043908A1 - Method and apparatus for multi-source dynamic network coding

Info

Publication number: WO2014043908A1
Application number: PCT/CN2012/081820
Authority: WO
Inventors: 李挥; 侯韩旭; 冯俊秋; 张华宇; 朱兵; 韩元波
Original assignee: 北京大学深圳研究生院; 深圳市矽伟智科技有限公司
Priority date: 2012-09-24
Filing date: 2012-09-24
Publication date: 2014-03-27
Also published as: CN104704760A; CN104704760B

Abstract

The present invention relates to construction of a multi-source dynamic network coding method, which comprises the following steps: preprocessing a source node, a receiving node and an intermediate node involved in a session; the source node sending identical data packets over all paths of the receiving node that currently requests data from the source node; sequentially determining the number of data packets that are from different sessions and transmitted on each link of the paths, setting a coding path, and notifying the coding path of the related source node; the source node selecting one or more data packets according to received coding path data and transmitting the data packet over multiple paths to the receiving node; and the intermediate node coding the received data packet. The present invention also relates to an apparatus for implementing the method. Implementing the multi-source dynamic network coding method and apparatus in the present invention has the following advantages: the coding complexity is lower and the data throughout will not be reduced.

Description

Multi-source dynamic network coding method and device

Technical field

The present invention relates to the field of network coding, and more particularly to a method and apparatus for multi-source dynamic network coding.

Background technique

Network Coding (NC, Network Coding), which was born in 2000, proposes how to achieve the limit of network communication capacity determined by the "maximum flow minimum cut theorem" when a single source is multicast or broadcast to multiple receiving points in a communication network. .

The routing switch on the traditional communication network node only performs the "store-and-forward" function, and regards the information as "goods", so that the information is considered to be non-superimposable. This cognitive limitation makes it difficult for the network to achieve maximum streaming. The NC indicates that if the routing switch is allowed to encode the input multiple streams of information before sending, the network communication can be maximized. This type of network coding is called intra-session network coding. For this type of network coding, the data output by the intermediate node is a linear combination of the input data. The receiving node can decode when it receives enough independent coded packets. However, the use of network coding for multiple sources in the network has always been a challenge. This type of network coding is called multi-source network coding, also known as multi-session network. Encoding problems or inter-session networks give coding problems.

When multiple multicast sessions are transmitted over the network, a practical approach is to allocate a subset of the network capacity for each session. If the sub-capacity allocated by each session satisfies the maximum flow minimum cut condition, the intra-session network coding can reach the limit of the maximum flow of the network. However, for general multi-source network coding problems, inter-session network coding is likely to reach the maximum network throughput. In general, inter-session network coding is more complicated than intra-session network coding. In order to ensure that each receiving node can decode, the encoding should be more careful, and the received packets cannot be linearly combined because the receiving node is very likely. There is not enough input capacity to solve the encoded package.

At present, for the general multi-source network coding problem, there is no practical network coding method with the best throughput. [N. Ratnakar, R. Koetter and T. Ho, "Linear flow equations for network coding in the multiple unicast case," in Proc. DIMACS Working Group Network Coding, Jan. 2005.] introduced the poison-antidote method for multiple unicast sessions. [S. Katti H. Rahul, W. Hu, D. Katabi, M. Medard, and J. Crowcroft, "XORs in the air: Practical wireless network coding," in Proc. ACM SIGCOMM, Sep. 2006.] An opportunistic XOR coding (OXC) method applied in a wireless network. Subsequently, the literature [M. Kim, M Medard, U. O'Reilly, and D. Traskov, "An evolutionary approach to inter-session network coding," in Proc. IEEE INFOCOM, Brazil, Apr. 2009.] proposed the choice The method of Selective Random Linear Coding (SRLC), in which the intermediate node uses selective random linear coding, and the coding algorithm is controlled by a genetic algorithm. Literature [Abdallah

Khrei shah, Chih- Chun Wang and Ness B. Shroff, "Rate Control With Pairwise

Intersession Network Coding", in IEEE/ACM Trans. Networking, JUNE 2010.] proposed two pairs of inter-session network coding (PINC), PINC only allows encoding of two data packets, coding use Random linear network coding. However, none of the above articles consider the dynamic changes of the network.

For the above-mentioned poison-antidote method (PA), the PA tracks all sessions. When a node encodes a packet, the coded packet is called a poison, and the packet used for decoding is called an antidote. Every time a poison is produced, an antidote must be sent immediately; the PA uses an XOR encoding method. The disadvantage of PA is its high complexity, which has a complexity of at least 0 (^ 1 £ 11 V I), where m is the number of sessions, k is the number of sessions allowed to encode, and does not take into account dynamic changes in the network.

For the above selection of random linear coding SRLC: The intermediate node uses selective random linear coding, and the coding coefficients are controlled by genetic algorithms. Its shortcomings ^^ because the algorithm will become more complicated as the number of sessions increases, making the SRLC algorithm very complex, and does not consider the dynamic changes of the network.

For the above two inter-session network coding algorithm PINC: PINC only allows encoding two data packets, the coding uses random linear network coding. The disadvantage is that the complexity is high, and only the encoding between the two sessions is allowed, and the throughput is reduced.

In addition, for the above-described opportunistic XOR coding OXC, the XOR coding method is used, and the computational complexity is low. The disadvantage is that the neighbor nodes of the coding node need to decode immediately, the throughput is reduced, and it is only suitable for application in the wireless network. Summary of the invention

The technical problem to be solved by the present invention is to provide a multi-source dynamic network with low complexity and low data throughput, which is high in complexity of the above-mentioned coding and may reduce data throughput. Coding method and device.

The technical solution adopted by the present invention to solve the technical problem is: Constructing a multi-source dynamic network coding method, comprising the following steps:

A) preprocessing the source node, the receiving node, and the intermediate node involved in the session for the plurality of sessions existing on the current network, so that the source node involved in the session obtains the path of the current session. a parameter that causes the tail node of the plurality of links involved in the session to obtain its link parameters;

B) said source node transmitting the same data packet on all paths of a receiving node for which data is currently requested;

C) sequentially determining the number of data packets transmitted from different sessions on each link in the path, such as only one data packet, setting a path related to the link in the session to a non-encoding path; The number of data packets transmitted from different sessions on the road is | |, where | | >1, then for a receiving node, set the |R _w | - i paths in different sessions involved in the receiving node to be encoded a path, the remaining path is a non-encoded path, and notifying the source node involved in the encoded path;

D) the source node selects one or more data packets to be transmitted on multiple paths to the receiving node according to the encoded path data received by the source node;

E) The intermediate node encodes and transmits the data packets sent by different source nodes it receives to the receiving node.

Further, the step A) further includes:

A1) searching for a data transmission path from the source node to the receiving node according to the data request that the receiving node sends to the source node; and storing the found path set on the source node ;

A2) according to the obtained path, in turn, the receiving node reversely routes to the source node, obtains multiple links included in each path, and stores the link parameters on the tail node of the link; The link parameters include: a source node involved, a receiving node, a number of paths, and a path number of the link. Further, the step A) further includes:

A3) Set all paths to non-encoded paths.

Further, the step C) further includes:

C1) judging whether the number of data packets |R _W | received from the different source nodes and transmitted to the same receiving node is greater than 1, if yes, performing step C3); otherwise, performing step C2);

C2) setting the path of the link from the source node to the receiving node as a non-encoding path, and jumping to step C4);

C3) setting ||-1 of the path from the different source node to the receiving node where the link is located is an encoding path, and the remaining one is a non-encoding path, notifying the source node, and performing steps C4);

C4) determining whether all paths from the source node to the receiving node in the session are set, and if so, performing step D); otherwise, repeating steps CI) - C4 for the remaining intermediate nodes.

Further, the step D) further includes: if there is no coding path in all paths of the source node to the receiving node, the source node sends the same data packet on all paths; If there are _A coding paths in all the paths from the source node to the receiving node, and the number of paths from the source node to the receiving node is greater than the number of the encoded paths, then the source node is in all |R (IR - / ^.) packets are sent simultaneously on each path.

Further, if the session changes, the steps A) - E) are performed on the changed receiving node and the source node requesting the data.

Further, the intermediate node XORs the data packets from different source nodes and transmits the encoded data packets on the link.

The invention also relates to an apparatus for implementing the above method, comprising:

a pre-processing unit: configured to perform pre-processing on a source node, a receiving node, and an intermediate node involved in the session for multiple sessions existing on the current network, so that the source node involved in the session obtains the current The path parameter of the session, and the tail node of the multiple links involved in the session obtains its link parameters;

An initial stage data packet sending unit: configured to send the same data packet on all paths of a receiving node to which the source node currently requests data; The encoding path setting unit is configured to sequentially determine the number of data packets transmitted from different sessions on each link in the path, for example, if there is only one data packet, setting a path involving the link in the session to a non-encoding path If the number of data packets transmitted from different sessions on the link is |R _W |, where | | >1, then for a receiving node, set |R _w in different sessions involved by the receiving node; | - i paths are coded paths, the remaining path is a non-coded path, and the source node involved in the coded path is notified; a packet transmitting unit is configured to cause the source node to receive the encoded path according to the source path Data, selecting one or more data packets to be respectively transmitted on multiple paths to the receiving node;

Coding unit: configured to cause the intermediate node to encode and transmit a data packet sent by a different source node that it receives to the receiving node.

Further, the pre-processing unit further includes:

a forward path search module: configured to search for a data transmission path from the source node to the receiving node according to a data request that is sent by the receiving node to the source node; and on the source node Store the set of found paths;

The reverse path searching module is configured to: in turn, reversely route the receiving node to the source node according to the obtained path, obtain multiple links included in each path, and store the link on the tail node of the link. The link parameter includes: a source node, a receiving node, a number of paths, and a path number of the link;

Initial Path Setup Module: Used to set all paths to non-coded paths at the beginning. Further, the encoding path setting unit further includes:

a link receiving data packet judging module: configured to determine whether the number of data packets |R _W | received from the different source nodes and transmitted to the same receiving node is greater than 1;

The non-coding path setting module is configured to: when the data packet received by the link is 1, set a path where the link is located, and the path from the source node to the receiving node is a non-encoding path;

An encoding path setting module: configured to set, in the path of the different source node to the receiving node, the encoding path, and the remaining one is a non-encoding path, and notify the source node;

a setting completion judgment module: configured to determine, by the source node to the receiving node in the session Whether all paths are set.

The method and apparatus for implementing the multi-source dynamic network coding of the present invention have the following beneficial effects: instead of performing network coding on all input data of a node, only performing network coding on the required data and using an exclusive OR coding method, The coding is less complex and does not reduce data throughput.

DRAWINGS

1 is a flowchart of a method in a method and apparatus for multi-source dynamic network coding according to the present invention; FIG. 2 is a specific flowchart of a pre-processing step in the embodiment;

Figure 3 is a flow chart showing the setting of the encoding path in the embodiment;

4 is a schematic comparison of performance of a multi-source dynamic network coding method and an existing coding technique in the embodiment;

Figure 5 is a schematic view showing a topology of the embodiment;

Figure 6 is a schematic view showing another topology of the embodiment;

Figure 7 is a schematic structural view of the device in the embodiment;

detailed description

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

As shown in FIG. 1, in the method and apparatus embodiment of the multi-source dynamic network coding of the present invention, the multi-source dynamic network coding method includes the following steps:

Step S11 performs pre-processing on the path between each source node and the receiving node according to the current data request: For the convenience of the following description, it is necessary to make a detailed description of the network environment in this embodiment. In the present embodiment, the communication network is represented by a directed network graph G = (V, E), where V represents a set of nodes in the directed graph and E is a set of links of the directed graph G. Let the capacity of all links in G be unity. The set of source nodes to the network is {s ₁ s ₂ , s _h }, the set of receiving nodes is {t ₁ t ₂ , and the set of receiving nodes of the data source Si is Ti, ί=1 , 2, ..., h, the multicast session is represented as ( _Si , Ti). There are a total of h multicast sessions in Figure G, i=l, 2, ..., h. Each multicast session (s _i T is represented by |Ti| unicast sessions (s _i tj), 7;, · = 1,2,...,Ι7; ΐ. And each session (S _i tj ) are composed of several paths, using R _{u to} represent the path set of the session (s _i tj), ={^| l≤k≤|R _u |}, |Ri,j| is the number of elements of the set, |R _U | is also the maximum flow of the session (s _i tj). ^ is the kth path of Ri,j, and the path is a directed from the starting point to the ending point tj Link, expressed as r = _Si ... tj. Each receiving node tj is described by three parameters (S(tj), L(tj), Rj). Where S(tj) is a set of source nodes requested by the receiving node tj, and a path set from the source set S(tj) to the receiving node tj is represented by Rj,

The set of links used by the receiving node tj when receiving the source is L(tj). The set of receiving nodes that are using link e is denoted by T _e . Tail(e) indicates the tail node of the directed link e, and the number of sessions transmitted by link e is expressed as). Since the capacity of all links is unit capacity, any link does not transmit data or transmits a data amount per unit time.

Each intermediate node records some identification information, which is information that the source node uses its downlink and information that the source requested by the receiving node uses its link. All intermediate nodes only need to assign appropriate coding packets to their downlinks based on these identification information. The source node only needs to count the encoded information fed back from the intermediate node and allocate the packet on its path.

It can be seen from the above description that there are multiple sessions on the network, and these sessions may be in one or more multicasts. In this step, for the multiple sessions existing on the current network, the sources involved in one dialog respectively The node, the receiving node, and the intermediate node perform pre-processing, so that the source node involved in the session obtains the path parameter of the current session, and the tail node of the multiple links involved in the session obtains its link parameter. It is worth mentioning that, in this embodiment, each session is processed according to the method of step S11-step S15, but for the convenience of description, it is not repeated one by one.

Step S12: The source node transmits the same data packet on each path to the receiving node: in this step, since each of the source nodes obtains a path capable of reaching the receiving node to which the data request is sent in the foregoing preprocessing step Therefore, a source node sends the same data packet on all paths of a receiving node that currently requests data; likewise, each source node is able to reach a data request to which it is sent in this step. The same packet is sent on the path of the receiving node.

The intermediate node on the path of step S13 sets the coding path according to the data packet parameters received by the node, and reports the source path to the source node. In this embodiment, the number of coding paths is the number of coding paths in the path set R _ld . Expressed by p _y . ρ indicates whether the path is an encoded path, . =1 indicates that the path is an encoded path, and opposite ^ =0 indicates a non-encoded path. Ρ Determined by the intermediate node of the path, and the number of encoded paths of the path set R is counted by the source node, p = ∑ Pi „ ρ defaults to 0. If 1)1>1 data packet participates in coding, then each of the receiving nodes T _e selects ( \ _R(e) \-l ) paths as encoding paths, the other is non-encoding paths, and the number of encoding paths Feedback to the corresponding source node. For the receiving node, the encoded packet has 1) 1 data packet to participate in the encoding, so an additional encoding packet that is not linearly related to it is needed to solve the encoded packet. In this embodiment, the intermediate node selects (wl-1) paths as the encoding path and feeds back to the corresponding source node, and then the source node allocates (( _e )ll) to the receiving node that received the encoded packet. A linearly uncorrelated encoding packet so that the receiving node can decode. This is also the main idea of this embodiment. For each receiving node i.e, assume that when the link e participates in encoding the path belonging to the receiving node tj as r _h i=l, 2, 1 /^) 1, then 11^11(13⁄4) is selected as the non-encoding path. , i=l, 2,

1 ^) 1 , tj T _e , the other is the encoding path. That is to say, in this step, the number of data packets from different sessions transmitted on each link in the path is determined in turn, for example, if there is only one data packet, the path related to the link in the session is set to be non-coded. Path; if the number of data packets transmitted from different sessions on the link is | |, where | | >1, then for a receiving node, set |R _w in different sessions involved by the receiving node | -i paths are coded paths, and the remaining path is a non-coded path, and the source node involved in the coded path is notified; basically, in this step, the code path is set with a receiving node and sends data to The plurality of source nodes of the receiving node are related.

In this step, the intermediate node notifies whether the path to which the source node belongs is an encoding path, and the source node counts the number of encoding paths. The number of encoding paths can help the source node to allocate a data rate on its path to maximize the throughput of the receiving node. Step S14: The source node sends one or multiple data packets according to the encoding path parameter: In this step, a source node selects one or more data packets according to the encoded path data received by the source node respectively. The node transmits to multiple paths of a receiving node; in fact, in this embodiment, one source node may need to send the same or different data packets for different receiving nodes, so, in this step, similarly, The above one source node may repeat the above operations for different receiving nodes respectively. In this embodiment, when the session has IR l paths, where there are _j encoding paths, The receiving node is enabled to decode the encoded packet transmitted in the encoding path, and the source node selects a non-encoding path for each encoding path, and transmits the same data packet in the encoding path and the non-encoding path. Therefore, it is possible to transmit (IR l- _ΐ ) different data packets in up to IR paths.

Step S15 encodes and transmits the data packet on the encoding path: In this step, each intermediate node encodes the data packet sent by the different source node it receives and passes through the link where the link is located and where the link is located. The path obtained in the above step is transmitted to the receiving node.

It is worth mentioning that, in this embodiment, the data requirements of the receiving node to the source node are not static. In some cases, for example, if a newly added data request or a previous data request is completed, The path condition of the above network changes. In this case, only the above steps need to be performed for the changed part, and the data in the still existing path and the source node, the receiving node, and the intermediate node are not needed. Or re-acquire after the parameter is cleared. This allows for a smaller cost or overhead to be used when the network changes, again maximizing network throughput.

As shown in FIG. 2, in the embodiment, the foregoing pre-processing step further includes the following steps: Step S21: Forward path search: In this step, path search of all current sessions is first performed, that is, according to each receiving node Or a data request submitted by a plurality of source nodes, respectively searching for a data transmission path of the source node to the receiving node by the one (in the case of multiple source nodes); and storing on the source node The set of paths found.

Step S22 performs reverse routing one by one according to the found path: In this step, according to the path obtained in the foregoing step, one receiving node sequentially reversely routes to the source source node to obtain multiple links included in each path. And storing the link parameter on the tail node of each link; wherein the link parameters include: a cell node involved, a receiving node, a number of paths, and a path number of the link. In this step, if the network includes multiple receiving nodes (or multiple receiving nodes that make data requests), each receiving node will perform this step separately.

Step S23 sets the initial state of all the paths: In this step, all the paths in the above session are set to be non-encoded paths due to the initialization requirements.

In preprocessing, it gets the tag information that the output is the session (s _i tj) at the intermediate node. Basically, in this embodiment, starting from the source Si, the information is routed according to the directed graph, and all the achievable by Si is traversed. Nodes and links, the nodes that pass through are marked as Si. From the receiving node tj, each link connected to the receiving node tj is reversely routed to the source node Si according to the information of the previous node being marked as _Si , and the tail(e) of all the links e passing through Each link e is labeled e(Si, tj, |R _M |, k), 1≤|<≤|[3⁄4|. In the node tail(e), each link e is marked as e(st _P |R _U | , k), l≤k<|R _M | , indicating that the path of the source _Si to be transmitted to the receiving node tj is shared by I Ri , j|, link e belongs to the kth path of the path set R _u . In this embodiment, the overall path of the receiving node is divided into three parts: The first part is a set of discrete paths, and the set of paths is represented. The second part is a non-discrete path set and all receiving nodes can decode, using (3⁄4 to represent its path set. The third part is a non-discrete path set and does not satisfy all receiving nodes can decode, called this part of the path set is not available Decoding the path set, the path set is represented by 13⁄4. For the first part of the path set, the non-encoding method can achieve the best throughput; the network coding can achieve the best of the second part of the path set throughput; In the three-part path set, encoding does not improve throughput. In this embodiment, a non-encoded routing strategy is used to process this part of the path set, because network coding can not only improve throughput but also generate a large amount of delay.

As shown in FIG. 3, in this embodiment, the setting of the encoding path further includes the following steps: Step S31: The number of data packets received by the link is greater than 1, if yes, step S32 is performed; if not, step S33 is performed; In this step, it is determined whether the number of data packets received from a different source node and transmitted to the same receiving node is greater than 1, and if so, step S32 is performed; otherwise, step S33 is performed; When there are multiple links, multiple links will perform the above operations separately.

Step S32: Set one of the paths from the different source nodes to the current receiving node of the link to be a non-encoding path, and the rest are encoding paths: In this step, set the different source nodes where the link is located. The path to the receiving node (the total number of these paths is |R _W |) is the encoding path, and the remaining one is the non-encoding path. After performing this step, step S34 is performed.

Step S33: setting the path where the link is a non-encoded path: In this step, since there is only one data packet on the link, the data packet is not required to be encoded, so the path where the link is located is a non-encoded path. . After performing this step, go to step S35.

Step S34: transmitting the information of the encoding path to the source node: In this step, the information of the encoding node set in the above step S32 is sent to the relevant source node, that is, the data packet is sent and passed through the chain. The path is transmitted to the source node of the above receiving node.

Step S35: The path of all the links is set: In this step, it is determined whether all the links (of course, the link related to the current data transmission) are both judged and set, and if so, jump The process proceeds to step S14; otherwise, step S36 is performed.

Step S36 is ready to judge the next link: In this step, it is prepared to judge and set the next unset link, and after performing this step, return to step S31.

In summary, in this embodiment, the network coding uses an exclusive OR operation instead of random network coding. In multi-source network coding, the data packets participating in the coding are all from different source nodes, and the probability that the same receiving node receives the same coded packet is extremely small, so the correlation of the coefficients is not considered. Even if the receiving node receives two identical encoded packets, the encoding node only needs to add a bit and then XOR code to the data participating in the encoding to ensure that the received two encoded packets are irrelevant. For example, two data packets _ai ( a^! a^ ... a!, _B ) and a ₂ ( a _{2; 1} a ₂ , ₂ ... a ₂ , _B ), the length of the data packet is B bits. If an encoding packet received by the receiving node is ai ^ a, the encoding process of another encoding packet is as follows: adding a bit Ό ' at the end of the packet a 々 header and a ₂ respectively to obtain a data packet (Oawau ...a !, _B ) and (a _{2 1} a ₂ , ₂ ... a ₂ , _B 0 ) and XOR the two data packets. The two encoding packets aea? and (Oa^au ... aw ) Θ ( a _{2 1} a ₂ , ₂ ... a ₂ , _B 0 ) are independent of each other, and the data packet a^ a ₂ can be decoded. In this embodiment, it is first necessary to initiate preprocessing for each pair of sessions (s _i tj). The source node transmits ( | RiJ - Pu) different data packets in its path | RiJ according to the number of encoded paths in its path set, and the intermediate node selects the appropriate data packet according to the source information requested by the receiving node and uses XOR coding, then selecting (R( _e )-1) paths for each receiving node T _e as the encoding path and the other as the non-encoding path, and finally feeding the encoding information back to the corresponding source node. In this way, the source node and the intermediate node in the network constitute a dynamic multi-source network coding algorithm that interacts with each other. Table 1 gives the meaning of the symbol representation in this embodiment. Table 1. Symbolic meaning.

In(v _k ) node v _k receives the original packet received by the packet node v _k in one transmission slot within one transmission slot

T is using the set of receiving nodes of the link v _k v _k+1 s (t receiving the path set of the source set session (S tj ) requested by the node tj

R(v _k v _k+ i) set of data packets to be transmitted in link v _k v _k+1

(gate)) The intersection of the source node packets received by the receiving node _{ν ν +ι}

Furthermore, when the action of the encoding process is described by the node itself, it can be seen that, for the receiving node, the main steps of the receiving node are:

Step 1: The receiving node tj requests the source node S (t packet, j=l, 2, ..., n.

Step 2: Start preprocessing for each pair of sessions (Si, tj), j=l, 2, ..., n, V^. e S(tj). For the source node, it can be roughly divided into an initialization phase and a recursive phase, wherein the actions of the initialization phase are: for the session (s _i tj), j=l, 2, ..., n, V^. e^ , let p _u =0, transmit the same packet on all paths of each session (s _i tj). The action in the recursive phase is: If |Ri,j| >Pi,j, transmits (|Rij| - Pi,j) packets on IRijl paths, and the encoding path in session (s _i tj) and one Non-coding path The same data packet is transmitted; if | |= , the same data packet is transmitted on |Rij| paths, and one encoding path is a non-encoding path, and only the original data packet is transmitted on the path. For the intermediate node, it can also be roughly divided into an initialization phase and a recursive phase, wherein the actions of the initialization phase are:

Step 1: For(k=l; k< | R _M | +1; k++) = 0. For(k=l; k<K; k++) = Φ. R(v _k v _k+1 )= kJ^-≡ In(v _k )f] ( Π^(^))). (Link v _{_k} v _k _{+ 1} packets received set of link v _{_k k} ₊ intersection set request packet receiving node ₁ v)

Step 2: Select (ΐ/?(ν _Λ+1 )ΐ-1) paths for each receiving node T _e as the encoding path (P =1) and the other path as the non-encoding path (= 0). The action of the recursive phase of the intermediate node is:

K

Step 1: (v _k v _k+1 )= {α,,Ια, ε/«(ν,)Π( gate S(t _q ) [j S ) ,

d ^=k+l

Step 2: Select ( \R(v _k v _k+1 )\-l ) paths for each receiving node T _e as the encoding path ( _P = 1) and the other path as the non-encoding path ( = ο). If the path ^ is an encoding path, the original data packet received by the node v _{k is} notified to the encoding node in the nodes _V1 , v ₂ , ... Vw. Therefore, when the path to which the intermediate node v _k belongs is an encoding path and v _k has multiple input links, the data packet S _q received by v _{k is} notified to the encoding node of the encoding path, and the encoding node can S _{q is} encoded without increasing the number of encoding paths because the data packet S _q in the encoded packet can be solved at the intermediate node V _k . In the initialization phase, the source node _Si transmits the same data packets in its path, which ensures that the receiving node can decode immediately. In the recursive phase, (IR - Pi, j) different data packets are transmitted according to the number of coding paths Pi, j. The intermediate node chooses to XOR the received data packet, (the gate ^( ) is the receiving node Γ _{ν ι} Find the same set of sources. And η( Γ^( )} is the data packet that needs to be transmitted on the link V _k V _k+ i in the initial stage. If Ι /^ _Λ+1 ) Ι =1 , then R(v _k v _k+1 ) is Raw data. Instead, an exclusive OR encoding operation is performed on {α., Ι α. G /«( n( fl ⁵ ^?))} and the encoded packet is transmitted in the link V _k V _k+ i . ϋ _d is the original packet set received by node v _d , so the uplink at node v _d can encode ϋ sd=k+ld=k+l without increasing the number of encoding paths. When the number of paths of the session (s _i tj) is equal to the number of encoded paths, even if the packets transmitted on all the paths are the same, the receiving node cannot decode, so only one number of encoding paths can be reduced, so that Making full use of the network resources can enable the receiving node to decode the encoded packet immediately, reducing the delay. In an actual network, the session requested by the receiving node changes over time. For this embodiment, only the receiving node that has requested the change needs to enter the initialization phase. After one transmission interval, the embodiment can enter the iterative phase. Note that in this embodiment, the iterative phase is the best phase. Both PA and SRLC require a large number of iterations to enter the optimal phase, and even enter the optimal phase within a reasonable time. In this embodiment, the number of transmissions of the source data packet, that is, the transmission rate, is adjusted according to the coding condition of the path in the network. So this embodiment is dynamically adaptive. In this embodiment, the coding algorithm is distributed in each intermediate node, which reduces the burden on the source node. It is a distributed multi-source network coding algorithm. The low complexity of the algorithm is mainly the space of the intermediate node. Computational ability in exchange for. In this embodiment, a GM NC (general multi-source network coding)-like multi-source network final coding algorithm is proposed. GM NC does not network encode all of the node's input data, but only performs network coding on the required data. The GM NC uses an XOR encoding method, and the conventional XOR encoding may reduce the throughput. An XOR encoding method proposed in this embodiment makes all XOR encoding packets linearly independent. In this embodiment, all the received packets received by the receiving node are solvable. Because if any receiving node receives an encoding packet +& ₂ +...+& _£ , it comes from g different source nodes.

S!, s ₂ , ..., s _g , according to the GMNC algorithm, the coding node will select g-1 paths as the coding path, such as the source SbSz, . . ^^ path, the corresponding g-1 letters The source node selects a path in (s^), i=l, 2, ..., gl, and sends the data packet ^, ., .Α^ respectively, and the receiving node can receive the data packet at the same time.

Therefore, for any receiving node tj, any encoded packets it receives @ a ₂ ten... ω 3⁄4 are solvable. In other words, GMNC can guarantee that all the encoded packets received by all receiving nodes can be decoded. Note that the traditional XOR operation in the finite field GF(2) may reduce the throughput of the network, in the packets aF^uau...a ₁₃ ) and a2=(a ₂ , ia ₂ , 2... a ₂ , _B ), respectively, 4 bbits '0' add force to the head and tail of ai and a ₂ , (Oauan Ha^) and (aa^H a^O), and then XOR, The coded packet is independent of the traditional XOR code package @ a ₂ . The intermediate node needs to record the tag information e(s ₁ tj, IRyl, k) for its downlink. There are up to hn sessions in the network, and all paths within each session are discrete, so each link e is used by up to hn paths. The entire network needs to store at most hnlEI tag information. Figure 4 shows the consumption of various coding strategies in different networks. For the poison-antidote method, we evaluated the consumption at k=3. For the selection of random linear network coding, each chromosome is composed of all coding vectors, so its consumption is at least ( . „, ^ 1^ 1 ^/ . In the network we evaluated, the number of nodes in the network is INI, Among them are INI/8 source nodes, INI/8 receiving nodes, each node has 5 uplinks and 5 downlinks. As can be seen from Figure 4, GMNC is superior to PA method in consumption performance. and SRLC, even for SRLC, the header needs to be stored in the coded coefficients g _m large _networks, §01 finite field GF (2 ^q) the elements in the finite field GF (2 ^q), the random linear network The error probability of coding is 1- (1-^2, where V is the maximum number of links in the network. In SRLC, each coded packet is encoded by at most h original packets, so the header of the SRLC is Hq bit. For a traditional XOR coded packet, the header size encoded by the hash packets is h. For g packets ( , & ₂ , g <h), in order to obtain two linear independent coding packets, a bit '0' needs to be added to the original data packet. One encoding package is _βι ® _β ₂ ®...® _β , and the encoding method of another encoding packet is as follows: Add the bit '0' to the head of a packet and the tail of g-1 packets, respectively, and then Or these g new packets. The two encoded packets obtained are independent of each other. For the second coded packet, the added redundancy bit is g, plus the coding coefficient of h bits, so the total header is (g+h). FIG. 5 is a topological structure of a network in this embodiment. In FIG. 5, the source S^PS ₂ is multicast to the receiving nodes ^ and t ₂ . The receiving node ^ has two sessions (s ₁ and (s ₂ , t _x ), the session (s ₁ has two paths _S1 t! and

The session (s^ti) has only one path 1 (^ 1 = 1. Similarly, the two sessions (s ₁ t ₂ ) and (s ₂ , t ₂ ) of the receiving node t ₂ also have one path and two respectively. Path, |R |=1, |R ₂ , ₂ |= 2. For the receiving node ^, the link _V1 V ₂ belongs to two paths of different sessions, so it is necessary to perform a network coding operation on the link _V1 V ₂ . For the receiving node t _l, the path s^v^ is the encoding path, and the other path s^v^ is the non-encoding path. The number of encoding paths is p ₁₁ = l and ρ ₂₁ =0 respectively. Similarly for the receiving node t ₂ , The number of coding paths is p ₂ , ₂ = l and ₁ , ₂ =0, respectively.

6 is another topology structure in this embodiment. In FIG. 6, three source nodes _S1 , s _2, and s ₃ are multicast to t ₂ and t _{3 respectively} , and multicast to ^, t ₂ , t ₃ , unicast to ^. That is, the source sets requested by the receiving nodes t _l t ₂ and t ₃ are respectively

S ₂ }. The capacity of the receiving node ^ to the source nodes _S1 , s ₂ , s ₃ is 3, the capacity of the receiving node t ₂ to the source nodes _S1 , s ₂ is 2, and the capacity of the receiving node t ₃ to the source node s ₂ is 2. . Therefore, in theory, the receiving nodes ^, t ₂ and t ₃ can respectively receive the requested source data packets 3, 2, and 2 in the unit transmission time. According to the GMNC algorithm, first find the discrete path of the respective session,

{ S ₂ V3V ₄ V5V ₆ t ₂ }, Ri, 3 ={ SiViV ₃ V ₄ V5V6t ₃ }, R ₂ , ₃ ={ S ₂ V3V ₄ V5V ₆ t ₃ , S ₂ V ₂ t ₃ }. Based on the tag information, the intermediate node can encode the received data packets and enable all receiving nodes to decode. In the initialization phase, all source nodes send the same packet on their path, and the intermediate node selects the appropriate packet encoding. In the initial phase, intermediate node v ₃ receives three packets, d, and R(v ₃ v ₄ ) = { a _l b! }. Therefore, the coded packet transmitted by the link ν ₃ ν ₄ is @ and the path Si V3V is selected for the receiving node ^, t ₂ , t ₃ , respectively. SiViV ₃ V ₄ V5V6t2, S ₂ V ₃ V ₄ V ₅ V ₆ t ₃ is the encoding path. When entering the iterative phase, the intermediate node V ₃ received three packets a,, d, and _{_{R (v 3 v 4) =}} {a 2, b 2, c 2}. Therefore, the coded packet transmitted by the link ν ₃ ν ₄ is a ₂ ten b ₂ ten c ₂ . Other nodes also send the appropriate data packets. Source _Sl, s ₂ and s ₃ packet are represented by a, b and c indicated in FIG. 6 by a, b, c respectively represent the source node s _l s ₂ and s ₃ packets of. The tag information of the intermediate node is as shown in Table 2. Table 2. All tag information for intermediate nodes.

Table 3. The throughput of Figure 3.

Table 3 gives a comparison of the throughput performance of the different coding strategies in Figure 6. In PINC, inter-session network coding is limited to special cases between two unicast sessions, so only data from source nodes Si and s ₂ or s ₂ and s ₃ can be on link v ₃ v ₄ The package is encoded. If the link v ₃ v ₄ transmits the encoded packets of a and b, the receiving node can not receive the data packet of the source s ₃ and can only receive the two data packets of the source _S1 and s ₂ , and the total throughput of the network. 6; if the link v ₃ v ₄ transmits the encoded packets of b and c, the receiving node t ₃ can only receive the data packet b, and the total throughput of the network is 6. Therefore, the maximum throughput that can be achieved by [5][10] is 6. The network coding algorithm in the session allocates a subtree for each source, and each source node transmits a corresponding data packet according to its own subtree, and the throughput of the network coding algorithm in the session is also 6. In FIG. 6, when the source node to the receiving unicast t _3, the source s ₂ to the receiving node unicast t _2, the source s ₃ ^ unicast to the receiving node. Obviously, GMNC algorithm may be such that each of three receiving node receives a data packet corresponding to the source, but can not satisfy the request PINC t ₃ at the same time request satisfies ^ and _12, that is, such that they can not be three The receiving nodes simultaneously receive a corresponding one of the requested source data packets. Therefore, compared to previous studies, the GMNC algorithm has great advantages in terms of total throughput and simultaneous requests for different receiving nodes. As shown in FIG. 7, in this embodiment, an apparatus for implementing the foregoing encoding method is further provided. The apparatus includes: a pre-processing unit 71, an initial stage data packet transmitting unit 72, an encoding path setting unit 73, and a data packet sending unit. 74 and coding unit 75; wherein the pre-processing unit 71 is configured to perform pre-processing on the source node, the receiving node, and the intermediate node involved in the session for the plurality of sessions existing on the current network, so that the message involved in the session is The source node obtains the path parameter of the current session, and causes the tail node of the multiple links involved in the session to obtain its link parameter; the initial stage data packet sending unit 72 is configured to request data from the source node at present. The same data packet is sent on all paths of a receiving node; the encoding path setting unit 73 is configured to sequentially determine the number of data packets transmitted from different sessions on each link in the path, for example, if there is only one data packet, the session is set. The path involved in the link is a non-encoded path; as the number of packets transmitted from different sessions on the link | | A, where, | |> 1, then for a receiving node, the receiving node to set different session according to the | R _W | -1 coding path is a path, a path of the remaining non-coding path, and Notifying the source node involved in the above encoding path; the data packet sending unit 74 is configured to enable the source node to select one or more data packets respectively in the multiple paths to the receiving node according to the encoded path data received by the source node. Up-conversion; the encoding unit 75 is configured to enable the intermediate node to encode and transmit the data packet sent by the different source node that it receives to the receiving node.

In this embodiment, the pre-processing unit 71 further includes a forward path search module 711, a reverse path search module 712, and an initial path setting module 713. The forward path search module 711 is configured to send the message according to the receiving node. a data request by the source node, respectively searching for a data transmission path from the source node to the receiving node; and storing the found path set on the source node; the reverse path searching module 712 is configured to obtain a path, in which the receiving node reversely routes to the source node, obtains multiple links included in each path, and stores the link parameters on a tail node of the link; the link parameter The method includes: a source node involved, a receiving node, a number of paths, and a path number of the link; and an initial path setting module 713 is configured to set all paths to non-encoded paths at the beginning. In the present embodiment, the encoding path setting unit 73 further includes a link receiving data packet determining module 731, a non-coding path setting module 732, an encoding path setting module 733, and a setting completion determining module 734; wherein the link receives the data packet The determining module 731 is configured to determine whether the number of data packets |R _W | received from the different source nodes and transmitted to the same receiving node is greater than 1; the non-coding path setting module 732 is used to determine whether the number of data packets received from the different source nodes and transmitted to the same receiving node is greater than 1; When the data packet received by the link is 1, the path from the source node to the receiving node where the link is located is a non-encoding path; the encoding path setting module 733 is configured to set the link. The |R _W | -1 in the path from the different source node to the receiving node is the encoding path, and the remaining one is the non-encoding path, and the source node is notified; the setting completion determining module 734 is used for judging Whether all paths from the source node to the receiving node in the session are set.

It should be noted that, in this embodiment, each unit and module in the foregoing apparatus may be disposed on the same physical carrier, or may be disposed on different physical carriers. For example, in the embodiment, the above-described encoding path setting unit is disposed on the intermediate node, and the pre-processing unit is disposed on the source node. However, it is not to be construed as limiting the scope of the invention. It should be noted that a number of variations and modifications may be made by those skilled in the art without departing from the spirit and scope of the invention. Therefore, the scope of the invention should be determined by the appended claims.

Claims

claims

1. A multi-source dynamic network coding method, characterized by including the following steps:

A) For multiple sessions currently existing on the network, preprocess the source node, receiving node and intermediate node involved in one of the sessions, so that the source node involved in the session obtains the current path of the session. parameters, and enable the tail nodes of the multiple links involved in the session to obtain their link parameters;

B) The source node sends the same data packet on all paths of a receiving node currently requesting data;

C) Determine the number of data packets from different sessions transmitted on each link in the path in turn. If there is only one data packet, set the path involving the link in the session to a non-coding path; as in the link The number _of data packets transmitted from different sessions on the road is | Coding path, the remaining one path is a non-coding path, and the source node involved in the coding path is notified;

D) The source node selects one or more data packets to transmit on multiple paths to the receiving node based on the encoding path data it receives;

E) The intermediate node encodes the data packets received by it from different source nodes and transmits them to the receiving node.

2. The method of multi-source dynamic network coding according to claim 1, characterized in that said step A) further includes:

A1) According to the data request made by the receiving node to the source node, respectively search for the data transmission path from the source node to the receiving node; and store the found path set on the source node ;

A2) Based on the obtained path, reverse routing is performed from the receiving node to the source node in sequence, multiple links included in each path are obtained, and the link parameters are stored on the tail node of the link; The link parameters include: the involved source node, the receiving node, the number of paths, and the path number where the link is located.

3. The method of multi-source dynamic network coding according to claim 2, characterized in that said step A) further includes:

A3) Set all paths to non-encoded paths.

4. The multi-source dynamic network coding method according to claim 3, characterized in that step C) further includes:

C1) Determine whether the number || of data packets received by the link from different source nodes and transmitted to the same receiving node is greater than 1, if so, perform step C3); otherwise, perform step C2);

C2) Set the path from the source node to the receiving node where the link is located as a non-coding path, and jump to step C4);

C3) Set |R _W | -1 of the paths from different source nodes to the receiving node where the link is located as a coded path, and the remaining one as a non-coded path, notify the source node, and Execute step C4);

C4) Determine whether all paths from the source node to the receiving node in the session have been set up. If so, perform step D); otherwise, repeat steps CI)-C4) for the remaining intermediate nodes.

5. The multi-source dynamic network coding method according to claim 4, wherein step D) further includes: if there is no coding path in all paths from the source node to the receiving node, Then the source node sends mutually different data packets on all paths; For example, there are _A coding paths in all paths from the source node to the receiving node, and the path from the source node to the receiving node When the number is greater than the number of coding paths, the source node sends data packets simultaneously on all |R _W | paths.

6. The multi-source dynamic network coding method according to any one of claims 1 to 5, characterized in that if the session changes, the changed receiving node and its source node requesting data are executed. Described steps A)-E).

7. The multi-source dynamic network coding method according to any one of claims 1-5, characterized in that the intermediate node performs XOR encoding on data packets from different source nodes and encodes the encoded data Packets are transmitted on this link.

8. A device for implementing the multi-source dynamic network coding method as claimed in claim 1, characterized in that it includes:

Preprocessing unit: used to preprocess the source node, receiving node and intermediate node involved in one conversation respectively for multiple conversations currently existing on the network, so that the source node involved in the conversation obtains the current information. path parameters of the session, and enable the tail nodes of multiple links involved in the session to obtain its link parameters;

Initial stage data packet sending unit: used to send the same data packet on all paths of a receiving node for which the source node currently requests data;

Encoding path setting unit: used to determine the number of data packets from different sessions transmitted on each link in the path in turn. If there is only one data packet, set the path involving the link in the session as a non-encoding path. ; If the number of data packets transmitted from different sessions on the link is |R _W |, where | | >1, then for all its receiving nodes, set |R in the different sessions involved in the receiving node _w | -i paths are coded paths, and the remaining one path is a non-coded path, and the source node involved in the coded path is notified; Data packet sending unit: used to enable the source node to send data according to the code it received Path data, select one or more data packets to be transmitted on multiple paths to the receiving node;

Encoding unit: used to enable the intermediate node to encode the received data packets sent by different source nodes and transmit them to the receiving node.

9. The device according to claim 8, characterized in that, the preprocessing unit further includes: a forward path search module: configured to respectively search for the data requested by the receiving node according to the data request made by the receiving node to the source node. The data transmission path from the source node to the receiving node; and storing the found path set on the source node;

Reverse path search module: used to perform reverse routing from the receiving node to the source node based on the obtained path, obtain multiple links included in each path, and store them on the tail node of the link. The link parameters; The link parameters include: the source node, the receiving node involved, the number of paths and the path number where the link is located;

Initial path setting module: used to set all paths to non-encoded paths at the beginning.

10. The device according to claim 9, characterized in that the encoding path setting unit further includes:

Link received data packet judgment module: used to judge whether the number |R _W | of data packets received by the link from different source nodes and transmitted to the same receiving node is greater than 1;

Non-coding path setting module: used to set the path from the source node to the receiving node where the link is located as a non-coding path when the data packet received by the link is 1; Coding path setting module: used to set one of the paths from different source nodes to the receiving node where the link is located as a coding path, and the remaining one as a non-coding path, and notify the source node;

Setting completion judgment module: used to judge whether all paths from the source node to the receiving node in the session have been set.