DETERMINISTIC DISTRIBUTED NETWORK CODING

0Associated
Cases 
0Associated
Defendants 
0Accused
Products 
0Forward
Citations 
0
Petitions 
1
Assignment
First Claim
13. 3. (canceled)
1 Assignment
0 Petitions
Accused Products
Abstract
A network and a communication method are described. The network comprises: source nodes, receiver nodes, and coding nodes. The coding nodes are connected with input links for communication of input signals to the coding nodes and output links for communication of output signals from the coding nodes. The output signals are a linear combination of the input signals. The coefficients of the linear combination are deterministically chosen based on local information available locally at the coding node.
0 Citations
No References
No References
20 Claims
 13. 3. (canceled)
 4. A network comprising:
one or more source nodes; one or more receiver nodes; and coding nodes, allowing communication of source processes to each receiver node, the coding nodes being connected with input links for communication of input signals to the coding nodes and output links for communication of output signals from the coding nodes, wherein the output signals from the coding nodes are a linear combination of the input signals, a plurality of said coding nodes being configured to generate coefficients of the linear combination using local node numbers or alternatively local link numbers as inputs to a deterministic mapping from the local node numbers or alternatively from local link numbers to the coefficients of the linear combination, wherein the linear combination of processes transmitted from the one or more source nodes present in each signal in the network is specified as a vector of the coefficients, each coefficient corresponding to a process to be transmitted from one or more source nodes, and wherein said coding nodes are configured to update and to transmit the vector of coefficients by applying to the vector of coefficients linear combinations, wherein the linear combinations applied to the vector of coefficients are the same as the linear combinations applied to data transmitted through the network.  View Dependent Claims (5, 6, 7, 8, 9, 10, 11)
 12. A method for transmitting processes from one or more sources to each receiver of one or more receivers in a network, the method comprising:
providing coding nodes between the one or more sources and the one or more receivers; providing, for each coding node, input links for transmitting input signals to the coding node, and output links for transmitting output signals from the coding node, the output signals being a linear combination of the input signals; and generating at a plurality of said coding nodes coefficients of the linear combination using local node numbers or alternatively local link numbers as inputs to a deterministic mapping from the local node numbers or alternatively from local link numbers to the coefficients of the linear combination, wherein the linear combination of processes transmitted from the one or more source nodes present in each signal in the network is specified as a vector of the coefficients, each coefficient corresponding to a process to be transmitted from the one or more source nodes, and wherein the vector of coefficients is transmitted through the network and updated at said coding nodes by applying to the vector of coefficients linear combinations, wherein the linear combinations applied to the vector of coefficients are the same as the linear combinations applied to data transmitted through the network.  View Dependent Claims (13, 14, 15, 16, 17, 18, 19)
 20. A network comprising:
one or more source nodes; one or more receiver nodes; and coding nodes, allowing communication of source processes to each receiver node, the coding nodes being connected with input links for communication of input signals to the coding nodes and output links for communication of output signals from the coding nodes, wherein the output signals from the coding nodes are a linear combination of the input signals, a plurality of said coding nodes being configured to generate coefficients of the linear combination using local node numbers or alternatively local link numbers as inputs to a deterministic mapping from the local node numbers or alternatively from local link numbers to the coefficients of the linear combination, wherein said mapping is a modular operation, said coding nodes being configured to generate said coefficients as a linear combination mod q of the local node numbers or alternatively of the local link numbers, wherein q is a constant, the linear combination of processes transmitted from the one or more source nodes present in each signal in the network is specified as a vector of the coefficients, each coefficient corresponding to a process to be transmitted from one or more source nodes, and wherein said coding nodes are configured to update and to transmit the vector of coefficients by applying to the vector of coefficients linear combinations, wherein the linear combinations applied to the vector of coefficients are the same as the linear combinations applied to data transmitted through the network.
1 Specification
The present application is a Continuation of U.S. application Ser. No. 15/958,178 filed Apr. 20, 2018, which, in turn, is a Continuation of U.S. application Ser. No. 15/609,856 filed May 31, 2017, now U.S. Pat. No. 9,979,666 issued on May 22, 2018, which, in turn, is a Continuation of U.S. application Ser. No. 13/940,703 filed on Jul. 12, 2013, now U.S. Pat. No. 9,699,104 issued on Jul. 4, 2017, which, in turn, claims priority to U.S. Provisional Application No. 61/672,208, filed on Jul. 16, 2012, all of which are incorporated herein by reference in their entirety. Furthermore, the present application may be related to U.S. Pat. No. 7,706,365, which is incorporated herein by reference in its entirety.
Nodes in a computer network can also function as encoders. In particular, a node operating as encoder does not just forward (i.e. relay or replicate) information received from an input link, it also encodes such information. Coding at a node in a network is known as network coding.
Networks can be modeled as graphs with unit capacity directed links, one or more discrete sources, and one or more receivers, as shown in R. Koetter and M. Medard, “An algebraic approach to network coding,” IEEE/ACM Transactions on Networking, Vol 11 Issue 5, October 2003, incorporated herein by reference in its entirety. A common communication problem on networks is the multicast connection problem, where all source processes have to be transmitted to each of the receiver nodes.
Network coding enables connections that are not possible if limited to forwarding. In particular, it has been shown in R. Ahlswede, N. Cai, S. Y. R. Li, and R. W. Yeung, “Network Information Flow”, IEEE Transactions on Information Theory and vol. 46, pp. 12041216 (2000), incorporated herein by reference in its entirety, that it is, in general, not optimal to simply route or replicate the information to be multicast. Rather, by employing coding at the nodes, bandwidth can generally be saved.
A network can be represented as a directed graph, as shown in R. Koetter and M. Medard, “Beyond Routing: An Algebraic Approach to Network Coding”, Proceedings of the 2002 IEEE Infocom (2002), incorporated herein by reference in its entirety. The graph comprises source nodes and receiver nodes, where discrete independent random processes (source processes) are observable at one or more of the source nodes and output processes are observable at the receiver nodes. In the above reference, an algorithm for finding a linear coding solution to a given multicast connection problem, using knowledge of the entire network topology, is disclosed.
However, in applications where communication is limited or expensive, it may be preferable to determine each node'"'"'s behavior in a distributed manner. Determination of node behavior in a distributed manner is based on information available locally at each node and/or minimal control signaling, without requiring centralized coordination or knowledge of the overall network topology.
It has been shown in T. Ho, R. Koetter, M. Médard, D. R. Karger and M. Effros, “The Benefits of Coding over Routing in a Randomized Setting,” International Symposium on Information Theory (ISIT) 2003, incorporated herein by reference in its entirety, that the multicast connection problem can be solved in a distributed manner by means of random linear network coding.
It has been shown in T. Ho, S. Jaggi, S. Vyetrenko and L. Xia, “Universal and Robust Distributed Network Codes,” Infocom 2011, incorporated herein by reference in its entirety, that the multicast connection problem can be solved with zeroerror deterministic distributed network codes, but with impractical complexity and large delay.
The present disclosure provides systems and methods to solve the multicast connection problem in a distributed manner by means of deterministic linear coding without requiring centralized coordination or knowledge of network topology.
According to an embodiment of the present disclosure, a network is provided. The network comprises: one or more source nodes, wherein source processes are observable at the source nodes; one or more receiver nodes, wherein receiver processes are observable at the receiver nodes; and coding nodes, allowing communication of the source processes to each receiver node of the receiver nodes, the coding nodes being connected with input links for communication of input signals to the coding nodes and output links for communication of output signals from the coding nodes, wherein the output signals are a linear combination of the input signals. In particular, the coefficients of the linear combination (the coding coefficients) are deterministically chosen based on local information available locally at the coding node. By way of example, the coefficients are chosen from a code alphabet (the set of possible values of the coding coefficients) such as a finite or infinite ring or field.
According to another embodiment a network is provided. The network comprises: one or more source nodes, wherein source processes are observable at the source nodes; one or more receiver nodes, wherein receiver processes are observable at the receiver nodes; and coding nodes, allowing communication of the source processes to each receiver node, the coding nodes being connected with input links for communication of input signals to the coding nodes and output links for communication of output signals from the coding nodes, wherein the output signals from a coding node are a linear combination of the input signals and wherein coefficients of the linear combination are deterministically chosen based on an adjustable mapping through a deterministic function.
According to a further embodiment, a communication method is provided. The method transmits processes from one or more sources to each receiver of one or more receivers in a network. The method comprises: providing coding nodes between the one or more sources and the one or more receivers; providing, for each coding node, input links for transmitting input signals to the coding node, and output links for transmitting output signals from the coding nodes, the output signals being a linear combination of the input signals; wherein an overall linear combination of the processes transmitted from the one or more sources present in each signal in the network is specified as a vector of coefficients, each coefficient being deterministically chosen based on local information available locally at the coding node and corresponding to a process to be transmitted from one or more sources, and wherein the vector of coefficients is transmitted through the network and updated at each coding node by applying to the vector of coefficients linear combinations, wherein the linear combinations applied to the vector of coefficients are the same as the linear combinations applied to data transmitted through the network.
According to further embodiments of the present disclosure, in case of suboptimal choices, the coefficients can be adjusted using feedback from the receiver nodes to improve the rate of information delivery.
Therefore, unlike the zeroerror codes in the above mentioned T. Ho, S. Jaggi, S. Vyetrenko and L. Xia, “Universal and Robust Distributed Network Codes,” Infocom 2011 paper, where the choice of coding coefficients entails complexity and delay that scales exponentially in the number of network nodes, the embodiments of the present disclosure achieve complexity and delay that scales only polynomially in the number of network nodes, by allowing for some possibility of error in the initial deterministic choice of coefficients and using feedback to improve the choice of coefficients.
In particular, prior deterministic network code designs are complex since they restrict their choice of coding coefficients to cases that guarantee success in delivering the information at maximum rate to its intended receivers. However, making such a guarantee without centralized knowledge of the network topology entails high complexity. On the other hand, the embodiments according to the present disclosure differ from these prior designs by removing the a priori guarantee of maximum rate information delivery. By allowing coefficients to be chosen at each node without the abovedescribed requirement of correctness, coefficient choice is greatly simplified. In particular, if the initial deterministic choice is suboptimal, feedback can be used to improve the rate through the use of bidirectional or reverse links.
Since the approach according to the embodiments of the present disclosure does not depend on the network topology, it can be operated in a distributed fashion, and is robust to changes in network topology, including addition or removal of links, nodes, sources or receivers.
The approach according to the present disclosure is useful in environments where networks are adopted, such as computer communications networks (especially overlay networks, ad hoc networks, or sensor networks), and distributed computer systems.
A first embodiment of the present disclosure concerns network coding over acyclic subgraphs of networks. Such acyclic subgraphs can be constructed using, for example, the techniques in D. S. Lun, N. Ratnakar, M. Médard, R. Koetter, D. R. Karger, T. Ho, and E. Ahmed, “MinimumCost Multicast over Coded Packet Networks,” IEEE Transactions on Information Theory, 52 (6), pp. 26082623, June 2006, incorporated herein by reference in its entirety. Data from the sources is transmitted over the acyclic subgraph to the receivers. If the network has bidirectional links or additional links in the reverse direction from the receiver nodes to intermediate nodes in the subgraph, feedback from the receivers can be used to improve the rate of information delivery.
The present disclosure considers approaches in which network nodes deterministically choose coding coefficients in a distributed fashion and transmit on each outgoing link a linear combination of incoming signals, specified by a vector of coding coefficients. The resulting signal is a linear combination of the source processes, so reference will be made to signals and vectors interchangeably in the following. To avoid high complexity, in accordance with several embodiments of the present disclosure, the deterministic coefficient choices are designed to achieve optimal rates in some network topologies. Suboptimal coefficient choices can subsequently be identified and improved using feedback, if desired, through the presence of reverse and/or bidirectional links in the network.
While any deterministic choice of coefficients can be used to obtain a suboptimal solution, not necessarily based on characteristics of nodes and/or links (e.g. using the same deterministic mapping for nodes and/or links and then possibly using feedback to improve the mapping), some properties are generally beneficial in obtaining better rates over a broader range of network types and topologies. This can help to reduce the amount of feedback, or the amount of rate loss when feedback is not used. As an example, in case of nodes with multiple outgoing links, it is generally beneficial for the outgoing links to have linearly independent coding coefficients. Additionally, or in the alternative, it is generally beneficial for outgoing links from different nodes to have linearly independent coding coefficients, as much as feasible, subject to complexity and topology constraints. As another example, if the network experiences packet losses, it is generally beneficial for each outgoing packet from a node to have coding coefficients which are linearly independent over time from coding coefficients of a subsequent outgoing packet from the same node. Example embodiments are given in the following two paragraphs.
In some embodiments of the present disclosure each node is associated with a node number, which can be obtained from some physical characteristics associated to the node (e.g., its location coordinates, IP address, hardware address or serial number) or any arbitrary deterministic numbering of nodes (e.g., if there are N nodes, each node can be assigned an index number from 1 to N in any arbitrary deterministic order). The node numbers can be used as inputs to a deterministic mapping from local node numbers to coding coefficients. Reference can be made, for example, to
C=c_{ijk}A+c_{ihk}B, (1)
where c_{ijk }and c_{ihk }are deterministically chosen coefficients. By way of example,
c_{ijk}=(i+aj+bk)mod q, (2)
while
c_{ihk}=(i+ah+bk)mod q, (3)
where a and b are arbitrary deterministic constant elements from the code alphabet and q is the alphabet size (for a finite alphabet), or a suitably large constant. In other words, (2) and (3) are deterministic mappings from local node numbers i, h, j, k to coding coefficients. If the network experiences packet losses, the coding coefficients associated with each outgoing packet can be made to vary over time by replacing equations (2) and (3) with
c_{ijk}=(i+aj+bk+t)mod q, (2′)
and
c_{ihk}=(i+ah+bk+t)mod q, (3′)
respectively, where t can be a local counter or clock signal.
In other embodiments, each link is associated with a link number, which can be obtained from some physical characteristics associated to the link or any arbitrary deterministic numbering of links (analogous to the node numbers described in the paragraph above). The link index numbers can be used as inputs to deterministic functions that map local link numbers to coding coefficients. Reference can be made, for example, to
C=c_{ij}A+c_{hj}B, (4)
where c_{ij }and c_{hj }are deterministically chosen coefficients. By way of example,
c_{ij}=(i+aj)mod q, (5)
while
c_{hj}=(h+aj)mod q, (6)
where a is an arbitrary deterministic constant element from the code alphabet and q is the alphabet size (for a finite alphabet), or a suitably large constant. In other words, (5) and (6) are deterministic mappings from local link numbers i, h, j to coding coefficients. Additionally, if the network experiences packet losses, equations (5) and (6) can be modified similarly to what done with equations (2) and (3) above.
In other embodiments, the deterministic functions mapping node or link numbers onto coding coefficients (see, e.g., equations (2), (3), (5) and (6)) can be influenced by feedback from receiver nodes, if bidirectional or reverse links are available. Reference can be made, for example, to
In other embodiments, not all nodes employ coding. As one example, if a node has enough available capacity to forward all its received information to each of its next hop neighbors, it does not need to do coding. As another example, a network may comprise some nodes that are equipped with coding capability and others that are not; the latter nodes just do forwarding. In another example scenario, a standard routing algorithm such as I. Kou, G. Markowsky and L. Berman, “A Fast Algorithm for Steiner Trees,” Acta Informatica, volume 15, number 2, 1981, is used to choose coefficients corresponding to a noncoded (routing) solution. Network coding is then introduced at a limited number of network nodes based on feedback from receiver nodes on their received rates. An example of such a scheme is given in the following paragraph. Such embodiments can reduce coding overhead in situations where coding is of no or limited benefit.
In one of such embodiments, shown in
In other embodiments, only some nodes employ deterministic coding while other nodes may employ existing randomized coding techniques (such as those shown in the above mentioned U.S. Pat. No. 7,706,365) or no coding.
The management information comprised of the various linear combinations can be maintained and sent through the network, for each signal in the network, as a vector of scalar coefficients for each of the source processes, and updated at each coding node by applying the same linear combinations or mappings to the coefficient vectors as to the data or information signals.
The above embodiments can be generalized to vector linear coding, by considering each link as multiple parallel links of smaller capacity, and considering each source as multiple colocated sources of lower rate.
The above embodiments can also be generalized to convolutional network coding over subgraphs with cycles, by considering coding coefficients as polynomials in a delay variable.
The above embodiments can also be generalized to network codes in which different portions of the subgraph from source to sink employ linear network coding over different alphabets. In some embodiments, the alphabet becomes larger with the distance from the source. In other embodiments, one or more intermediate nodes decode messages and reencode using a different field.
The methods and systems described in the present disclosure may be implemented in hardware, software, firmware or combination thereof. Features described as blocks, modules or components may be implemented together (e.g., in a logic device such as an integrated logic device) or separately (e.g., as separate connected logic devices). The software portion of the methods of the present disclosure may comprise a computerreadable medium which comprises instructions that, when executed, perform, at least in part, the described methods. The computerreadable medium may comprise, for example, a random access memory (RAM) and/or a readonly memory (ROM). The instructions may be executed by a processor (e.g., a digital signal processor (DSP), an application specific integrated circuit (ASIC), or a field programmable logic array (FPGA)) By way of example and not of limitation, as shown in
While several illustrative embodiments of the invention have been shown and described in the above description, numerous variations and alternative embodiments will occur to those skilled in the art. Such variations and alternative embodiments are contemplated, and can be made without departing from the scope of the invention as defined in the appended claims.