Method and apparatus for data transmission

By collecting path availability information between the source node and the destination node, determining the path cost and using network decoding technology, dynamic path management of the multi-path transmission protocol is realized, and the problem of low scheduling and path management efficiency in the prior art is solved. Improves network throughput and reliability.

CN115486042BActive Publication Date: 2024-11-15CODE ON NETWORK CODING LLC
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202180030575.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Priority Date
2020-04-23
Filing Date
2021-04-23
Publication Date
2024-11-15
Estimated Expiration
2041-04-23

AI Technical Summary

Technical Problem

Existing multipath transmission protocols, such as MPTCP, have limitations in scheduling and path management, especially when dividing data across multiple paths and dynamically adjusting scheduling, it is difficult to effectively utilize multiple available paths to improve network throughput and reliability. .

Method used

By collecting path availability information between the source node and the destination node, determining the cost factors of each path, and using this information to allocate data units and select paths, redundant degrees of freedom (DoF) is generated using network decoding technology to achieve Multipath transmission.

Benefits of technology

Improve the throughput and reliability of multipath communication, dynamically adjust path allocation to deal with network changes, and reduce data transmission delay and resource waste.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115486042B_ABST
    Figure CN115486042B_ABST
Patent Text Reader

Abstract

In some embodiments, a method includes: transmitting a first data unit from a source node to a destination node via a first path; receiving information about the availability of one or more second paths between the source node and the destination node; determining a cost associated with each of the one or more second paths; allocating a first number of second data units to the first path and allocating a second number of second data units to the one or more second paths based on the determined costs; and transmitting the second data unit from the source node to the destination node via the first path and / or the one or more second paths according to the allocation, wherein the first data unit, the second data unit, or both the first and second data units include network-coded data.
Need to check novelty before this filing date? Find Prior Art

Description

Background Art

[0001] As is known in the art, several attempts have been made to use multiple paths for transmission protocols from a source node to a destination node in a computer network. An example of a multipath protocol is the Multipath Transmission Control Protocol (MPTCP). MPTCP is one of several multipath protocol implementations proposed by the Internet Engineering Task Force (IETF) working group.

[0002] Implementations of multipath protocols such as MPTCP can potentially improve the way current computer networks operate by increasing the reliability and throughput of data communication sessions using the principle of resource pooling. For example, if a communication path between a source and a destination fails or becomes highly congested, data can be sent via an alternative path, thereby avoiding any interruption of the data communication session.

[0003] In addition, multipath protocols potentially enable clients (e.g., mobile devices) to establish connections simultaneously through multiple network interfaces. Modern communication devices (e.g., smart phones and other handheld communicators, tablet computers with communication functions, etc.) generally have multiple network interfaces for use with different network technologies with different connection characteristics. Such an interface can provide a connection with a cellular network, a wired and wireless local area network (LAN) or other devices in a self-organizing or peer-to-peer mode. For example, different network technologies may all have different delays, throughputs, and reliability characteristics. For example, a mobile device connected to a remote source node by a cellular network and an IEEE 802.11 wireless LAN can observe different usage costs and service quality through corresponding interfaces.

[0004] The goal of a multipath protocol such as MPTCP can be to utilize multiple available paths and network resources to perform communications, increase throughput, reduce data transfer time, and / or efficiently utilize available resources. With the advent of more powerful (in terms of computing power and communication interfaces) devices, expectations for such benefits have increased. Existing multipath protocols and implementations are limited in one or more aspects. For example, one problem in MPTCP is scheduling. The scheduling problem in MPTCP involves partitioning data across multiple paths and dynamically readjusting the schedule based on congestion feedback.

[0005] The use of coding (in general), and specifically random linear network coding (RLNC), has been proposed to alleviate the scheduling problem of multipath communications. U.S. Patent No. 9,877,265 ("the '265 patent") describes techniques for facilitating multipath transmission by implementing a transport protocol that emulates the current Transmission Control Protocol (TCP) / MPTCP interface (to higher and lower layers) and using coding to provide more efficient service. U.S. Patent No. 9,537,759 ("the '759 patent") describes the use of RLNC in a manner that utilizes traditional communication techniques and protocols, providing robustness to link failures (particularly in wireless networks), and the ability to roam from one network to another. For example, the '759 patent describes an embodiment that can operate on traditional transport layer protocols (e.g., TCP). Both the '265 and '759 patents are incorporated herein by reference in their entirety.

[0006] As is known in the art, there are transport layer protocols designed to operate over UDP (User Datagram Protocol). One such protocol is QUIC, which is primarily designed to reduce latency by: (1) eliminating head-of-line blocking seen by the Transmission Control Protocol (TCP) by establishing a set of multiplexed connections between endpoints, (2) improved congestion estimation, and (3) inclusion of a form of forward error correction (FEC). The term "QUIC" is sometimes used as an acronym for "QuicK UDP Internet Connections", but the IETF uses "QUIC" as the protocol name, not the acronym.

[0007] It has been proposed to extend QUIC to multipath operation. For example, see Q. De Coninck, O. Bonaventure, "Multipath Extension for QUIC," Internet Draft, October 30, 2017 (hereinafter referred to as "DeConinck"). The subsequent information Internet draft summarizes a set of recommended multipath requirements for QUIC, see C. Huitema, "QUIC Multipath Requirements," Internet Draft, January 7, 2018 (hereinafter referred to as "Huitema"). De Coninck proposed a congestion control mechanism that is managed separately at the path level, in which a handshake is provided to establish a multipath connection. As suggested in Huitema, De Coninck allows path-independent confirmations (ie, feedback) to be transmitted between end hosts (ie, source nodes and destination nodes). De Coninck proposes additional protocol features, such as path update frames, path state management, and confirmation formats, and provides for the collection of path statistics by end hosts for flow control purposes. Both De Coninck and Huitema are incorporated herein by reference in their entirety. Summary of the invention

[0008] According to one aspect of the present disclosure, in a system having a source node and a destination node, a method may include: transmitting a first data unit from the source node to the destination node via a first path; receiving information about one or more second paths between the source node and the destination node, the received information including information about path availability or path requirements; determining a cost associated with the first path and each of the one or more second paths; based on the determined cost, determining a first allocation of a first number of second data units to the first path and a second number of the second data units to the one or more second paths; and transmitting the second data unit from the source node to the destination node via the first path and via the one or more second paths, or via the first path and the one or more second paths according to the first allocation. The first data unit, or the second data unit, or both the first data unit and the second data unit may include network-coded data.

[0009] In some embodiments, the first number of second data units and the second number of second data units are both greater than 0. Transmitting the second data unit from the source node to the destination node may include: transmitting the second data unit from the source node to the destination node via the first path and the one or more second paths according to the first allocation.

[0010] In some embodiments, receiving information about the availability of the one or more second paths may include receiving information in a data request sent by the destination node. In some embodiments, the method may include, in response to receiving information about the availability of one or more second paths between the source node and the destination node: generating redundant degrees of freedom (DoF) by applying network decoding to the second data unit; and transmitting the redundant DoF from the source node to the destination node through the first path and the one or more second paths according to the first allocation.

[0011] In some embodiments, determining the cost associated with each of the one or more second paths may include determining one or more cost factors associated with each of the one or more second paths. In some embodiments, the one or more cost factors may include at least one of: an absolute cost factor associated with a path; a relative cost factor of two different paths; or a total cost factor associated with two or more paths. In some embodiments, the one or more cost factors may include at least one of: line delay; usage cost; load; congestion; or line variability. In some embodiments, determining the one or more cost factors associated with each of the one or more second paths may include: measuring packet delay; or receiving path status information from a network management unit.

[0012] In some embodiments, determining the cost associated with the first path and each of the one or more second paths may include: using a cost control function based on at least one of: a fixed budget; a time-varying budget; a fixed budget and a bandwidth usage prediction; a time-varying budget and a bandwidth usage prediction; or a bandwidth usage prediction. In some embodiments, the number of second data units allocated to the one or more second paths may be inversely proportional to the determined cost associated with the one or more second paths.

[0013] In some embodiments, the method may include: receiving updated information about one or more second paths between the source node and the destination node; determining an updated cost associated with each of the one or more second paths; determining a second allocation of a first number of third data units to the first paths and a second number of the third data units to the one or more second paths based on the determined updated costs; and transmitting the third data units from the source node to the destination node via the first path and the one or more second paths according to the second allocation. In some embodiments, the method may include, in response to determining the updated cost associated with each of the one or more second paths, terminating at least one of the second paths. In some embodiments, the method may include, in response to determining the updated cost associated with each of the one or more second paths, changing a coding scheme, coding redundancy, or one or more of a plurality of degrees of freedom (DoF) used when transmitting the third data units from the source node to the destination node.

[0014] In some embodiments, the method may include: transmitting multiple data units from the source node to the one or more intermediate nodes via multiple paths between the source node and the one or more intermediate nodes, wherein the one or more intermediate nodes are configured to transmit the encoded data units to the destination node via one or more paths between the one or more intermediate nodes and the destination node.

[0015] In some embodiments, the one or more intermediate nodes may include one or more access points (APs), wherein at least a first path of the multiple paths between the source node and the one or more APs corresponds to a WiFi network, wherein at least a second path of the multiple paths between the source node and the one or more APs corresponds to a cellular network. In some embodiments, the one or more intermediate nodes may be configured to decode data units received from two or more different source nodes together.

[0016] According to another aspect of the present disclosure, a method may include: obtaining a plurality of original packets to be transmitted to a destination node at a source node; generating decoded packets by linearly combining two or more of the plurality of original packets using network decoding; determining a cost associated with transmitting data on each of a plurality of paths; and distributing one or more of the decoded packets among a plurality of paths leading from the source node to the destination node, wherein a total number of packets allocated to each of the one or more plurality of available paths is selected based on the determined cost associated with each path; and transmitting the one or more decoded packets to the destination node based on the distribution.

[0017] According to another aspect of the present disclosure, in a system having one or more source nodes, at least one intermediate node and a destination node, wherein the intermediate node and the destination node are capable of establishing a transmission session and transmitting encoded data between the intermediate node and the destination node via one or more available different output paths, a method may include: transmitting data destined for the destination node from the one or more source nodes to the intermediate node via one or more connections; encoding the data at the intermediate node; and transmitting the encoded data from the intermediate node to the destination via one or more connections.

[0018] According to another aspect of the present disclosure, a device for use at a source node to transmit data to a destination node may be provided. The device may include: a processor; and a non-volatile memory storing computer program code. When executed on the processor, the stored computer program code may cause the processor to perform a process that is operable to: transmit a first data unit from the source node to the destination node via a first path; receive information about one or more second paths between the source node and the destination node, the received information including information about path availability or path requirements; determine a cost associated with each of the first path and the one or more second paths; determine an allocation of a second data unit to the first path and to the one or more second paths based on the determined cost; and transmit the second data unit from the source node to the destination node via the first path, via the one or more second paths, or via the first path and the one or more second paths according to the determined allocation.

[0019] In some embodiments, receiving information about the availability of the one or more second paths may include receiving information in a data request sent by the destination node. In some embodiments, the device may include, in response to receiving information about the availability of one or more second paths between the source node and the destination node: generating redundant degrees of freedom (DoF) by applying network decoding to the second data unit; and transmitting the redundant DoF from the source node to the destination node through the first path and the one or more second paths according to the determined allocation.

[0020] In some embodiments, determining the cost associated with each of the one or more second paths may include determining one or more cost factors associated with each of the one or more second paths. In some embodiments, the one or more cost factors may include at least one of: an absolute cost factor associated with a path; a relative cost factor of two different paths; or a total cost factor associated with two or more paths. In some embodiments, the one or more cost factors may include at least one of: line delay; usage cost; load; congestion; or line variability. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate one or more embodiments of the present disclosure and, together with the description of the example embodiments, serve to explain the principles and implementations of the present disclosure.

[0022] Figure 1 is a network diagram illustrating multipath communications in a computer network according to an embodiment of the present disclosure.

[0023] Figure 2 is a flow chart of a process for opportunistic multipath communication according to an embodiment of the present disclosure.

[0024] Figure 3 is a network diagram illustrating how one or more nodes of a computer network may use opportunistic multipathing according to an embodiment of the present disclosure.

[0025] Figure 4 is a flow chart of a process for path splitting according to an embodiment of the present disclosure.

[0026] Figure 5A , Figure 5B and Figure 5C is a network diagram illustrating how one or more nodes of a computer network may use path splitting according to an embodiment of the present disclosure.

[0027] Figure 6 is a network diagram illustrating conversion between multiple different protocols according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0028] This article describes structures and techniques for use in multipath communication networks, including networks that use decoding to implement multipath operations at the transport layer. Various embodiments provide path management and path splitting in multipath communication applications. In some embodiments, an intermediate network device (IND) can be used as a virtual endpoint to partially terminate connections and convert between different transport protocols, including converting to or from a decoding multipath protocol. The structures and techniques sought to be protected herein can be applied to heterogeneous networks, including but not limited to mobile network applications.

[0029] Figure 1 Multipath communication in a computer network 100 according to some embodiments of the present disclosure is shown. A source (or "sender") 110 may be communicatively coupled to a destination (or "receiver") 160 via multiple paths 150a, 150b, etc. (generally 150). Paths 150 may correspond to logical associations between source 110 and destination 160. Multiple paths 150 may exist between a source and a destination. For example, a path from a video server to a smartphone may be established via different radio interfaces such as WiFi (IEEE 802.11) and LTE (Long Term Evolution). Figure 1 The example of includes two paths 150a, 150b.

[0030] Source 110 and destination 160 may correspond to end hosts (or "hosts") of transport operations to perform the multiplexing and encapsulation operations required at the transport layer. Figure 1 As shown, multiplexing may be performed by a multiplexer 130 at source 110 and a demultiplexer 170 at destination 160. Source 110 and destination 160 may also perform flow control and reliability operations.

[0031] A plurality of data streams 120a, 120b, etc. (generally 120) may be established between a source 110 and a destination 160. The data streams (or "streams") 120 may be used to transfer application data from the source 110 to the destination 160. Figure 1 In the example of , source 110 is shown as sending two data streams 120a, 120b to destination 160. Each stream 120 can be segmented or mapped into one or more frames (e.g., frame 122). Frames from different streams 120 can be multiplexed together and formed into packets (e.g., packet 144) and transmitted on one or more paths 150. Frames and packets can also carry control information.

[0032] One or more connections 140a, 140b, 140c, 140d, etc. (generally 140) may be established between source 110 and destination 160, where each connection is associated with a given path 150. Connections 140 may carry packets between source 110 and destination 160 and may be characterized by different properties such as error rate, latency, encryption requirements, reliability, etc. Figure 1 In the example of FIG. 1 , the first path 150a has one connection 140a, and the second path 150b has three connections 140b, 140c, 140d.

[0033] The flow of data in a transport protocol typically begins with a request from a destination 160, typically in the form of a number of octets (or bytes). Such requests typically follow a credit-based flow control protocol, examples of which include QUIC and Hypertext Transfer Protocol 2.0 (HTTPS / 2). In such protocols, a credit represents permission to transmit a given number of octets.

[0034] Transport protocols such as TCP and QUIC typically rely on sliding windows at both the source 110 and the destination 160 to track frames (or packets) sent and received. At the source 110, feedback in the form of packet acknowledgments (ACKs) from the destination 160 (along with transmission credits) enables the source 110 to remove frames from its window, allowing it to insert new frames from the stream, slide the window, and continue transmitting. The destination 160 inserts received frames into its window and typically removes them as they are sent in order to the stream requesting application.

[0035] A coded transport protocol (such as described in the '265 and '759 patents) performs coding of data units (e.g., frames, packets, or octets) to enable multipath operation and improve reliability. Such a coding protocol is capable of introducing redundant erasure data units (redundancy) at any node in the network (e.g., at a data "choke point"), and therefore, can be used to achieve a desired coding rate for a flow or connection.

[0036] The decoded transport protocol uses the concept of decoding degrees of freedom (DoF), where a DoF is a transmission data unit that is useful for a decoding operation. The term "degrees of freedom" (DoF) is used herein to represent both decoded or undecoded (i.e., native) data units. In RLNC, a decoded data unit is a linear combination of undecoded data units. A DoF window or decoding window can be used by a source 110 to control the amount of data it decodes together at any given time, and is used at a destination 160 to collect received packets and perform decoding operations, such as described in the '265 and '759 patents. A source 110 and a destination 160 can establish a connection 140 on one or more network paths 150, where the source 110 distributes frames among the available connections 140. The frames allocated to each path 150 can be undecoded or decoded. In either case, they can be considered degrees of freedom (DoF).

[0037] The network 100 may be controlled by the source 110, by the destination 160, by both the source 110 and the destination 160 in a cooperative manner, and / or by a network management unit 180 that is separate from the source 110 and the destination 160. The network management unit (or "network controller") 180 may be an independent node of the network 100, or may be part of another node in the network, such as a router, a switch, an access point (AP), a base station, a gateway, a middlebox, or a proxy node. Although embodiments of the present disclosure may be described as having the source 110 establish and manage paths and connections, it should be understood that similar functionality may be implemented within the destination 160 and / or the network management unit 180. In some embodiments, the network management unit 180 may instruct any two nodes along a given network path to split and terminate a new path / connection.

[0038] Still reference Figure 1 , the source 110 may include a connection manager 112 configured to manage paths and connections in an opportunistic manner based on a static or dynamic cost function, including path / connection establishment, path / connection termination, and modification of transmission characteristics across one or more paths / connections. A cost function may include any mathematical function that maps a value representing a cost to one or more communication parameters (e.g., bandwidth, utilization, or latency). The destination 160 may include a path monitor 162 to collect information about the characteristics (e.g., availability, latency) of various network paths 150a, 150b, etc. and the requirements of the data flow (e.g., requested rate, latency, jitter). The path monitor 162 may be further configured to measure the cost or one or more cost factors associated with each of the paths 150. The cost factors may include transmission metrics that are useful in determining the path cost.

[0039] Path cost may be determined by source 110 or destination 160 using various cost factors such as path delay (i.e., average or individual packet delay through a given path), path usage (i.e., additive or relative usage of a given path per transmitted data unit, as in the example above), path load (i.e., available load in frames per unit time or bits / second per available path), congestion (i.e., a measure of congestion on a given path, such as packet delays associated with congestion or the likelihood of delivery on any given path), or path variability (variability of any transmission metric, such as jitter or variation in packet reception times).

[0040] Path cost may be expressed as an absolute measure based on one or more cost factors, as a relative value (or "advantage"), or as a function of one or more absolute measures and relative values.

[0041] The path cost may be static during a session, or may vary dynamically during a session. In some embodiments, the path cost may vary in response to changes in one or more metrics measured by source 110 or destination 160, such as frame delay (i.e., any statistical information related to the time it takes for a frame to partially or completely reach a destination node). In some embodiments, the path cost may vary in response to information collected or generated by network management unit 180, such as information about the state of the path / connection, or statistical information related to packet delay, congestion, latency, and / or packet loss. In some embodiments, the network management unit may determine the cost and assign the path by communicating with source 110 and / or destination 160.

[0042] In one embodiment, the number of DoFs assigned to each path 150 may be selected such that they are related (e.g., according to a given mathematical relationship such as an inverse proportionality) to the cost incurred across each path 150. Figure 1As shown by dashed line 191 in (dashed line 192 represents the flow of data units from source 110 to destination 160), a connection 140 across path 150 can be established in response to a request for data sent from destination 160 to source 110. Connection 140 can be associated with a session (e.g., an application session) established between source 110 and destination 160. Data request 191 may include information about the availability, requirements, and cost factors of one or more paths 150 within network 100. Such path information can be generated by a path monitor 162 within destination 160. The transmission of cost and other path information from destination 160 to source 110 can be referred to as a "cost notification." In other embodiments, source 110 can measure / collect path information directly. For example, connection manager 112 can be configured to perform some or all of the processing described herein for path monitor 162. In still other embodiments, path information can be generated by an independent network management unit 180 and transmitted to source 110 and / or destination 160 via path 150 or via a separate network control channel.

[0043] After establishing multiple connections 140 across one or more paths 150, source 110 and destination 160 may exchange information about path costs via cost advertisements that may be carried by control frames. Source 110 (or more specifically, connection manager 112 at source 110) may then proceed to adjust path transmission characteristics (including possible path termination) based on the cost function. To make these adjustments, source 110 may use given targets (e.g., target maximum cost per unit time for operating a path, target maximum cost per unit time per session, etc.). Such targets may be communicated to source 110 by destination 160 or by network management unit 180 before or during a session.

[0044] As an example, destination 160 may correspond to a user device (e.g., a laptop, smartphone, or tablet) or an access point (AP) connected to both a WiFi network and an LTE network or other types of cellular networks. That is, path 150a may correspond to a WiFi network and path 150b may correspond to an LTE network. Fixed or variable costs may be associated with each path 150a, 150b. As previously described, one or more cost factors may be used to determine the cost. Assuming that the cost associated with LTE network 150b during a first time interval T1 is twice the cost associated with WiFi network 150a, the number of data units (e.g., frames, packets, or DoFs) assigned to (or "directed to") LTE network 150b during the first interval T1 may be set to half the number of data units assigned to WiFi network 150a. In some embodiments, a fixed or time-varying budget may be used to determine the number of data units assigned to path 150 and / or connection 140.

[0045] As used herein, the term "budget" refers to the total cost allowed per unit of time (e.g., interval) or per network element (e.g., flow, connection, path, or allocated for communication between source 110 and destination 160). In the above example, assume that (1) WiFi 150a and LTE 150b networks provide 100 Mbps and 50 Mbps, respectively, (2) the total bandwidth requested is 75 Mbps, (3) the cost of WiFi 150a and LTE 150b networks is 1 and 2 cost units per Mb, and (4) the budget allocated during communication is 2 cost units per second. Given the budget, the maximum bandwidth on LTE 150b is 50 Mbps, while WiFi 150a can support the required 75 Mbps at a cost of 0.75 cost units per second, which is within the budget. In the above example, factors other than bandwidth (e.g., packet delay, jitter, path variability) can be included in the cost function and budget allocation, thereby potentially optimizing communication parameters (e.g., rate) on the two paths.

[0046] In some embodiments, bandwidth usage predictions may be used to determine the number of data units assigned to paths 150 and / or connections 140. Such predictions may be based on historical path characteristics as well as more recent connection and / or flow statistics. Artificial intelligence and machine learning mechanisms may be used to improve usage predictions.

[0047] Upon receiving the cost notification, source 110 may take one or several courses of action. First, source 110 may terminate an existing path / connection or establish a new path / connection based on the path cost calculated at the source or destination. Second, source 110 may modify the decoding scheme across one or more paths / connections based on the calculated path cost. For example, source 110 may adjust the decoding redundancy or code rate based on the calculated path cost. This may include generating new decoded frames (i.e., redundancy or degrees of freedom) specifically for the additional path / connection, stopping the decoding operation, or stopping the transmission of decoded frames on a given path / connection based on the calculated path cost.

[0048] Figure 2 A process 200 for opportunistic multipath communication according to an embodiment of the present disclosure is shown. In the present disclosure, the term "opportunistic multipath communication" refers to communication between nodes of a computer network using multiple available paths between the nodes, where new paths can be utilized when they become available. Similarly, an "opportunistic multipath session" refers to a multipath session initiated in response to a new path becoming available (e.g., a single path session between a source and a destination can be converted to a multipath session in response to a new path becoming available between the source and the destination).

[0049] Portions of process 200 may be performed at a source (e.g., Figure 1 of source 110), destination (e.g., Figure 1 destination 160) and / or a network management unit (e.g., Figure 1 The network management unit 180 is implemented within and executed by them.

[0050] At block 201, a source may detect that a new path to a destination is available. This may be performed prior to initiating a session or while a session is pending. In some embodiments, block 201 may include determining that a new network interface has become available. It is understood that a path may become available in response to changing network conditions (e.g., a path becomes available after recovery from congestion or failure) or due to the dynamic nature of the network (e.g., a source moves within range of a new access point).

[0051] In response to detecting that a new path is available, at block 202, path information may be collected and used to evaluate the path costs of the new path and existing paths. It should be understood that there may not be an existing path initially, and therefore, the new path may correspond to the first available path. Figure 1 As discussed in more detail, the collected path information may include information regarding path delay, path usage, path load, path congestion, and / or path variability information.

[0052] In block 203, the path cost determined in step 202 may be used to determine the distribution of data units (e.g., frames, packets, or DoFs) over an updated set of available network paths (e.g., one or more available network paths). For a given path, the distribution of data units may be defined by the rate at which data units are transmitted by the source on the path. The number of data units allocated to a given path may be zero or greater than zero. In some embodiments, block 203 may include determining by the source whether to terminate one or more connections or paths within the network. Termination of a path refers to the termination of a logical association between the source 110 and the destination 160, and may include the termination of all connections that use the path or are otherwise associated with the path. For example, a given path may be terminated due to persistently high costs or upon receiving an indication from a network management unit.

[0053] At block 204, the source may transmit the data unit to the destination over one or more paths according to the distribution determined in 203. Process 200 may be repeated in whole or in part when new paths and / or new path information are available to the source, the destination, and / or the network management unit. For example, the source may apply the distribution of the data unit determined at block 203 until new path information becomes available to the source, at which point process 200 may be repeated from block 202. As another example, if a new path becomes available between the source and the destination, process 200 may be repeated from block 201.

[0054] In the WiFi / LTE example discussed above, a device running a 3Mbps connection over WiFi at a cost of $5 / Mbps can initiate a new path through an LTE interface at a cost of $10 / Mbps. For purposes of illustration, assuming that the goal is to send as much data as possible via LTE with a total budget of less than $20 / Mbps, process 200 can calculate that the bandwidth usage for WiFi and LTE should be 2Mbps and 1Mbps, respectively. Further assuming that equal-sized data units are used, process 200 can determine that for every two data units sent over WiFi (e.g., using a simple polling process), one data unit should be sent over LTE. The present disclosure is not limited to any particular path management process. A device can update its path information when it receives new information about an LTE path (e.g., network management, channel state information, packet loss, etc.). As shown above, such an update may result in a new determination of the DoF distribution across paths. With respect to the structures and techniques for which protection is sought herein, a variety of cost functions and methods for distributing data units across network paths may be implemented and used.

[0055] Figure 3 Describes how to employ opportunistic multipathing (e.g., based on Figure 2 The process 200 of FIG. 200 is shown in FIG. 201 . A multipath session may be initiated at session initialization or during an ongoing session between a source and a target. In the present disclosure, a multipath session that is initiated when a path becomes available is referred to as an opportunistic multipath session. An opportunistic multipath session may be established between a source and a target at any time during a session, including at session initialization.

[0056] refer to Figure 3 In an embodiment of the present invention, a simplified network 300 may include a source 301 and a destination 302. Source 301 and destination 302 have an ongoing session using a first (or "primary") path 311. During the session, source 301 determines that a second path 312 to destination 302 has become available. Source 301 may collect information and determine the costs of paths 311 and 312, for example, using the above method for Figure 2 The source may determine the distribution of data units (eg, frames, packets, or DoFs) across the two paths 311, 312, for example, using the above for Figure 2 The source may transmit data units across the paths 311, 312 according to the determined distribution, and may terminate one or more paths 311, 312 in response to changing network conditions or path states.

[0057] The process of establishing a multipath session may depend on the underlying network technology (e.g., bandwidth allocation mechanism, media access control (MAC), centralized control vs. distributed control, etc.). It may be performed as follows: (a) by a network management unit that instructs any two nodes along the path to split and terminate a new path / connection; (b) at the source 301; and / or (c) at the destination 302. Figure 4 An illustrative process for path splitting is described in the context of .

[0058] For example, in wireless, wired, or hybrid networks, new paths may become available to carry a connection (e.g., a WiFi network becomes available at a device using LTE to download or upload content). In wireless networks, the use of a second access point (AP) or base station (cellular) may become possible or advantageous temporarily or permanently during a session. A multipath session may be initiated at a device, AP / base station, or gateway node. In addition, opportunistic multipath sessions may improve performance during soft handoff when more than one network is available to a device.

[0059] The methods described herein can be used opportunistically within a session. An opportunistic multipath session can be initiated by an application endpoint by opening a new socket or port to communicate with another endpoint (e.g., a web client to a server). Similarly, a router or switch can make a new port available, allowing adjacent nodes to dynamically detect and initiate new paths. In addition, a wired or wireless mesh network node can discover a new route between two nodes, allowing the mesh node to discover a new path. The use of network decoding facilitates multipath decoding because the use of DoF instead of packets simplifies the data transmission protocol (e.g., simplifies the in-order arrival requirement) and reduces the state information (e.g., metadata) required for reliable communication.

[0060] In more detail, network decoding facilitates the use of opportunistic multipathing because any available bandwidth on any path can be easily used to transmit DoF without the need for coordination between nodes, thereby reducing complexity and resource usage in the network. Using network decoding, any set of available paths can be used to transmit DoF to allow opportunistic path selection to achieve the desired goals, such as high throughput or low decoding delay. Without network decoding, scheduling would be required, and scheduling would typically be unable to opportunistically utilize multipath resources (e.g., because doing so might disrupt scheduling), or would only be able to use replication, which would severely limit the usefulness of opportunistic multipath resources. According to the disclosed embodiments, the decoding scheme does not need to be aware of the fact that opportunistic multipathing is being used. This is because, in the case of network decoding, throughput is primarily determined by DoF, and not necessarily by the way they are transmitted. Where the goal is low latency, the decoding scheme can be adapted to opportunistic multipathing in order to reduce latency characteristics.

[0061] refer to Figure 4 , a path split is the action of starting a new multipath session during a session between a source and a destination. Path splitting can occur in response to a set of triggers, including but not limited to the following: (a) discovery of a new path or route (e.g., a new Internet Protocol, IP, route); (b) discovery of a new node in a network (e.g., a new node in a mobile wireless mesh network); and / or (c) relative or absolute changes in the performance of an existing path (e.g., a primary path), including degradation of one path relative to another for a given performance metric or cost factor (e.g., congestion, frame loss rate, etc.). The performance of a path can be quantified using a cost function as described above.

[0062] Path splitting can be directional (i.e., specific to the downlink or uplink) or bidirectional (i.e., non-specific), depending on the network technology used. For example, in a cellular network, path splitting can be applied to the uplink from the device to the base station, but not to the downlink. In other words, the download can remain unchanged, while the uplink can be split to use multiple paths from the device to the base station (e.g., through cooperating devices of other base stations). In some embodiments, path splitting is not limited to a single opportunistic path. Instead, multiple additional paths can be opportunistically started or terminated during a session.

[0063] Figure 4 4 shows a process 400 for path splitting according to an embodiment of the present disclosure. The illustrative process 400 may be performed by one or more nodes of a computer network (e.g., by Figure 1 Source 110) is implemented and executed.

[0064] At block 410, the source may initialize one or more paths and one or more connections to prepare for a session with the destination. At block 411, the source receives a new data unit (e.g., a frame). The data unit may be received from a higher layer of the source network stack, such as from an application layer within the source node. The received data unit 411 may then be buffered 412 and decoded 413 to generate additional DoFs. The received data unit 411 may be sent to both processing blocks 412 and 414 in a parallel or simultaneous manner.

[0065] At block 414, the source may determine a distribution of data units to be sent across one or more paths associated with the session. The distribution may be determined using a cost function, as described above in Figure 2 The received data unit 411 may be decoded 413 to generate the required DoF for each path based on an initial assessment of the DoF distribution. That is, once the DoF distribution and rate 414 for each connection are determined, additional DoFs may be generated by decoding 413 and transmitted across one or more paths / connections based on their distribution 414.

[0066] The received data units may also be buffered (block 412) and decoded (block 413) to generate redundant DoFs to satisfy the DoF distribution. In particular, at block 412, the data units may be added to a fixed or variable size decoding buffer maintained by the source. At block 413, the data units within the decoding buffer may be combined / decoded together using RLNC or other types of network decoding.

[0067] In some embodiments, as shown in optional block 415, the source may perform decoding at the level of individual paths. In particular, the source may generate and send redundant DoFs for a particular path based on the determined DoF distribution (from block 414). After block 414 and / or block 415, the process may be repeated from block 411, as shown in optional block 415. Figure 4 shown.

[0068] The illustrative process 400 also includes blocks 420 - 423 , which may be performed in parallel with the previously described blocks 410 - 415 .

[0069] At block 420, a decision may be made to perform a path split between the source and the destination. As previously described, path splitting may occur in response to one or more triggers, such as the discovery of new paths / routes / nodes in the network or detected changes in the relative or absolute performance / cost of existing paths. The source may receive information about path performance and new path availability in a periodic or continuous manner.

[0070] At block 421, after a decision has been made to perform path splitting, the source may initialize one or more new paths. At block 422, the source may determine a new distribution of data units (e.g., DoFs) across paths associated with the session (including recently initialized paths). The new distribution may be based on a cost function and an associated cost factor for each path. The new distribution may then be used during multipath transmission of the DoFs to the destination (at block 414). At block 423, paths that are deemed unused by the source may be terminated. After block 423, processing may be repeated from block 420, as shown in FIG. Figure 4 shown.

[0071] according to Figure 4 4. The process shown, the source node can adjust the redundancy on each path. Buffering can be performed specifically for the purpose of decoding or re-decoding (as shown in blocks 413 and 415). For example, the N most recent data units (e.g., frames) can be buffered (block 412) for the purpose of decoding and balancing DoF between paths (block 414). Path-specific re-decoding can be performed for the purpose of compensating for path loss or further balancing between paths (block 415).

[0072] Figures 5A-5C Three path splitting scenarios involving the participation of intermediate nodes according to some embodiments of the present disclosure are shown. Figures 5A-5C Similar reference numerals are used in the drawings.

[0073] refer to Figure 5A , a simplified network 510 may include a source node 511, an intermediate node 512, and a destination node 513. Initially, a session between source 511 and destination 513 may include path 514 between source 511 and intermediate node 512 and path 515 between intermediate node 512 and destination 513. During the session, or when the session is established, a second path 516 between source 511 and intermediate node 512 may become available to the source. In response, a multipath session based on paths 514 and 516 may be established between source 511 and intermediate node 512. The session may be referred to as a "partial" multipath session because when there are multiple paths between source 511 and intermediate node 512, only a single path 515 may be available to destination 513 (e.g., Figure 5A ). In this example, source node 511 may be referred to as a split node because it is the node from which the auxiliary path branches out. Intermediate node 512 may be referred to as a merge or termination node because it is the node at which the auxiliary path 516 merges with the primary path 514.

[0074] like Figure 5BAs shown, an auxiliary path 525 may be established between an intermediate node 512 and a destination 513. In this example, the intermediate node 512 may be a split node and the destination 513 may be a merge node.

[0075] refer to Figure 5C , the network 530 may include a source node 531, a first intermediate node 532, a second intermediate node 533, and a destination node 534. Initially, a session between the source 531 and the destination 534 may include a path 535 between the source 531 and the first intermediate node 532, a path 536 between the first intermediate node 532 and the second intermediate node 533, and a path 537 between the second intermediate node 533 and the destination 534. During the session or at session establishment, a second path 538 between the first and second intermediate nodes 536, 538 may become available. In response, the first intermediate node 532 may perform a path split to establish a partial multipath session between the source 531 and the destination 534, such as Figure 5C In this example, the first intermediate node 532 may be a split node, and the second intermediate node 533 may be a merge node.

[0076] Partial multipath sessions can be managed using the same techniques described herein for "full" opportunistic multipath sessions. For opportunistic multipath sessions, the process of establishing a partial opportunistic multipath session depends on the underlying network technology and can be performed by network management or at a split / merge node. Additionally, path splitting can occur in response to a set of triggers that may include, but are not limited to, discovery of new routes, new nodes, or fluctuations in path performance.

[0077] Due to the participation of intermediate nodes, partial multipath sessions and the accompanying path splitting and path merging operations can be performed without knowing the source, destination, or both. It can be performed locally (e.g., in a distributed system) without network management intervention. In addition, path splitting and merging may only involve the endpoints of the new path.

[0078] In some embodiments, path splitting can be performed as a combinatorial process that can occur opportunistically and recursively on recently added paths independently of the remaining participating nodes. For example, an intermediate node that is part of a recently added path may become a split node for a new path.

[0079] The path splitting process at a given node may vary based on whether network decoding is used on the primary path. If decoding is used on the primary path, the splitting node can generate redundancy by re-decoding buffered frames (rather than creating a new decoding stage) and distributing them among the available paths. Path management can follow the procedures described above and in Figure 2 and Figure 4If decoding is not used on the primary path, the split node may start decoding the packets to distribute them among multiple paths. In some embodiments, the merge node may perform decoding before merging the frame with the undecoded frame. In both cases, the split node may adjust the DoF on all participating paths, such as Figure 2 and Figure 4 shown.

[0080] refer to Figure 6 According to some embodiments of the present disclosure, network 600 may include an intermediate network device (IND) 602 communicatively coupled between source 604 and destination 606. IND 602 ​​may be provided as a middlebox, proxy, or other type of device separate from source 604 and destination 606. In other embodiments, the IND functionality described herein may be implemented in whole or in part within source 604 and / or destination 606.

[0081] IND 602 ​​may perform conversions between multiple different protocols (e.g., decoded / undecoded TCP, decoded / undecoded QUIC, etc.). For a given connection, device 602 may act as a virtual endpoint, partially terminating a given undecoded connection (e.g., via TCP or QUIC) before encoding packets using network coding (NC) for transmission within a network using network coding. For example, source 604 may send undecoded data units (e.g., frames, packets, or DoF) to IND 602 ​​over a single path / connection 608 using conventional TCP. IND 602 ​​may receive and buffer those data units, perform network coding on the buffered data units, and then distribute the decoded data units to destination 606 over multiple paths 610, 612. In one example, paths 610 and 612 may represent a WiFi network and an LTE network, respectively.

[0082] The operation of IND 602 ​​may not be visible to source 604 and / or destination 606. For example, source 604 may communicate using TCP while destination 606 communicates using the decoded multipath QUIC protocol. From the source's perspective, the destination uses TCP, and from the destination's perspective, the source uses the decoded multipath QUIC protocol. IND 602 ​​converts between the two different protocols.

[0083] Conversion of a connection means decoding a data frame or packet received over one transport layer connection using a given protocol (e.g., QUIC, NC / TCP) and retransmitting the data in the frame or packet to the destination using the new protocol. Example conversions include QUIC to NC / QUIC, QUIC to TCP, NC / TCP to NC / QUIC, multipath NC QUIC to TCP, etc. The letters "NC" in these acronyms refer to network coding.

[0084] IND 602 ​​may be provided as a networking device located within network 600 (potentially between a closed LAN and WLAN) that may partially initiate and terminate incoming connections or paths from multiple sources and toward multiple destinations. IND 602 ​​may also be a receiver or destination of a connection.

[0085] IND 602 ​​may partially terminate a connection and act as a virtual endpoint for both the source and destination. In some embodiments, IND 602 ​​may combine frames from multiple streams received across multiple incoming paths into one or more new (potentially multipath) outgoing connections. This includes topologies with: (a) multiple connections / paths input and a single connection / path output, (b) multiple source inputs to a single path / connection output, (c) multiple source inputs to a multipath output where one or more incoming or outgoing connections / paths use transcoding. In the present disclosure, multi-source operation means transmitting one or more streams from more than one source simultaneously.

[0086] IND 602 ​​can adjust protocol parameters and operations based on observed link statistics, sender / receiver requirements, user programmable inputs, etc. IND 602 ​​can adjust protocol parameters and operations based on statistics and requirements observed from multiple connections across one or more sender-receiver pairs. Examples of protocol parameters that can be adjusted by IND 602 ​​include the size of the decoding buffer (or "decoding window"), flow / congestion control parameters, error correction parameters, encryption strength, path selection, the amount of redundancy generated by the decoding process, encryption parameters, the set of paths used to transmit data units, and the distribution of data units on these paths. Adjustable protocol operations can include the transmission of acknowledgments (ACKs) between management nodes, buffer size notifications, connection handshakes, and other control frames that may be required for successful protocol operations. IND 602 ​​can also adjust / manage protocol parameters based on sender / receiver QoS requirements. For example, for protocols using network decoding (NC), this can include selecting decoding window size and field size.

[0087] In further embodiments of the present disclosure, the IND may adjust the protocol based on statistics and requirements observed from multiple connections across one or more sender-receiver pairs. Protocol operations include management of ACKs, buffer size announcements, connection handshakes, and transmission of other control frames that may be required for successful protocol operation.

[0088] In one embodiment of the present disclosure, the IND may partially terminate (uncoded) transport protocol connections (e.g., QUIC, TCP) from multiple sources to the same destination, and encode frames of one or more data streams from them using the NC for transmission over one or more NC / TCP or NC / QUIC connections. In the present disclosure, partial termination of a connection at the IND means that the IND provides the transport and control plane mechanisms (e.g., feedback, data requests, etc.) of the destination node, effectively terminating the connection with the source while forwarding the stream data to the destination.

[0089] The IND may also partially terminate the decoded transport protocol (NC / TCP, NC / QUIC), decode the packets, and retransmit them over the (undecoded) transport protocol connection (e.g., QUIC, TCP). Some embodiments may include conversion between QUIC and the decoded multipath QUIC protocol.

[0090] For both the above-described decoded-to-undecoded and undecoded-to-decoded conversions, the IND may adjust the decoding scheme (eg, modify the code rate) of the decoding path according to the destination QoS or congestion requirements.

[0091] If both the above decoded to undecoded and undecoded to decoded conversions are based on QUIC (e.g., input and output IND connections use QUIC, MP / QUIC, and NC-MP / QUIC), then the IND can also encode / decode data from multiple streams together to enable streams sharing routing segments to share decoding resources. This scheme can potentially use NC to combine redundancy of streams from different sources.

[0092] The IND may select which protocol to use for its outgoing connection (e.g., conventional vs. NC / QUIC) based on sender / receiver characteristics and requirements and / or link QoS statistics. The IND may also partially terminate connections from multiple sources over different transport protocols (e.g., one source using TCP and one source using QUIC) and convert these connections into a single connection to the destination by using a transport protocol that uses NC and / or multipath (e.g., MPQUIC / NC). Multiple INDs may communicate with each other to manage multipath / multisource transports. For example, referring to Figure 5C , the intermediate nodes 532, 533 may include representative INDs.

[0093] The subject matter described herein may be implemented in digital electronic circuits or in computer software, firmware or hardware (including the structural units disclosed in this specification and their structural equivalents), or in a combination thereof. The subject matter described herein may be implemented as one or more computer program products, such as one or more computer programs tangibly embodied in an information carrier (e.g., in a machine-readable storage device) or embodied in a propagation signal, for a data processing device (e.g., a programmable processor, a computer or multiple computers) to execute or control its operation. A computer program (also referred to as a program, software, software application or code) may be written in any form of programming language, including a compiled or interpreted language, and it may be deployed in any form, including as a stand-alone program or as a module, component, subroutine or another unit suitable for use in a computing environment. A computer program does not necessarily correspond to a file. A program may be stored in a portion of a file storing other programs or data, in a single file dedicated to the program in question, or in multiple coordinated files (e.g., files storing one or more modules, subroutines or portions of code). A computer program may be deployed to execute on a single computer or multiple computers at a site or distributed across multiple sites and interconnected by a communication network.

[0094] The processes and logic flows described in this specification (including the method steps of the subject matter described herein) can be performed by one or more programmable processors that execute one or more computer programs to perform the functions of the subject matter described herein by operating on input data and generating output. The processes and logic flows can also be performed by special purpose logic circuits, and the apparatus of the subject matter described herein can be implemented as special purpose logic circuits, such as FPGAs (field programmable gate arrays) or ASICs (application specific integrated circuits).

[0095] Processors suitable for executing computer programs include, for example, both general and special purpose microprocessors, and any one or more processors of any type of digital computer. Typically, the processor will receive instructions and data from a read-only memory or a random access memory or both. The basic elements of a computer are a processor for executing instructions and one or more memory devices for storing instructions and data. Typically, a computer will also include one or more mass storage devices for storing data, such as magnetic disks, magneto-optical disks, or optical disks, or be operatively coupled to receive data from or transfer data to or both of the one or more mass storage devices. Information carriers suitable for embodying computer program instructions and data include all forms of non-volatile memory, including, for example, semiconductor memory devices, such as EPROM, EEPROM, flash memory devices, or magnetic disks. The processor and memory may be supplemented by or incorporated in a dedicated logic circuit.

[0096] It should be understood that the disclosed subject matter is not limited to its application to the details of the constructions set forth in the following description or shown in the accompanying drawings and the arrangement of components. The disclosed subject matter can have other embodiments and can be practiced and executed in various ways. In addition, it should be understood that the words and terms used herein are for descriptive purposes and should not be considered restrictive. Therefore, those skilled in the art will understand that the concepts on which the present disclosure is based can be easily used as the basis for designing other structures, methods and systems to perform multiple purposes of the disclosed subject matter. Therefore, the claims should be deemed to include such equivalent constructions as long as they do not deviate from the spirit and scope of the disclosed subject matter.

[0097] While the disclosed subject matter has been described and illustrated in the foregoing exemplary embodiments, it should be understood that the present disclosure is made by way of example only and that numerous changes may be made in the details of its implementation without departing from the spirit and scope of the disclosed subject matter.

Claims

1. A method for data transmission operating in a system having a first node and a second node, the method comprising: transmitting a first data unit from the first node to the second node via a first path; receiving, by the first node, the second node, or both, information about one or more second paths between the first node and the second node, the received information including information about path availability or path requirements; determining, by the first node, the second node, or both, a cost associated with the first path and each of the one or more second paths; transmitting a second data unit from the first node to the second node via the first path; determining, by the first node, the second node, or both, allocation of the second data unit to the one or more second paths based on the determined costs; Generating, by the first node, a network-coded data unit for transmission via the one or more second paths by: If network decoding is not used in the first path, decoding the second data unit, and If network coding is used in the first path, re-coding the second data unit; and transmitting the network-coded data unit from the first node to the second node via the second path.

2. The method of claim 1, wherein: Receiving information regarding availability of the one or more second paths includes receiving the information in a data request sent by the second node.

3. The method of claim 1, comprising: In response to receiving information regarding availability of one or more second paths between the first node and the second node: generating redundant degrees of freedom DoF by applying the network decoding to the second data unit; as well as The redundant DoF is transmitted from the first node to the second node via the first path and the one or more second paths according to the determined allocation.

4. The method of claim 1, wherein: Determining a cost associated with each of the one or more second paths includes determining one or more cost factors associated with each of the one or more second paths.

5. The method of claim 4, wherein: The one or more cost factors include at least one of the following: the absolute cost factor associated with the path; the relative cost factors of two different paths; or The total cost factor associated with two or more paths.

6. The method of claim 4, wherein: The one or more cost factors include at least one of the following: Line delays; Cost of use; load; congestion; or Line variability, and Wherein, determining the one or more cost factors associated with each of the one or more second paths comprises: Measuring packet delay; or Receive path status information from the network management unit.

7. The method of claim 1, wherein: Determining a cost associated with the first path and each of the one or more second paths includes: Use a cost control function based on at least one of the following: Fixed budget; time-varying budgets; Fixed budget and bandwidth usage forecast; Time-varying budgets and bandwidth usage forecasts; or Bandwidth usage forecast.

8. The method of claim 1, wherein: The number of second data units allocated to the one or more second paths is inversely proportional to the determined costs associated with the one or more second paths.

9. The method of claim 1, comprising: receiving updated information about one or more second paths between the first node and the second node; determining an updated cost associated with each of the one or more second paths; determining a second allocation of third data units to the one or more second paths based on the updated costs; transmitting the second allocation of the third data unit from the first node to the second node; as well as The second network-coded data unit is generated by: If network decoding is not used in the first path, decoding the third data unit, and If network decoding is used in the first path, re-decoding the third data unit; and transmitting the second network-coded data unit from the first node to the second node via the one or more second paths.

10. The method of claim 9, comprising: In response to determining an updated cost associated with each of the one or more second paths, at least one of the second paths is terminated.

11. The method of claim 9, comprising: In response to determining an updated cost associated with each of the one or more second paths, changing one or more of: a coding scheme, coding redundancy, or a number of degrees of freedom (DoF) used when transmitting the third data unit from the first node to the second node.

12. The method of claim 1, comprising: transmitting a plurality of data units from the first node to the one or more third nodes via a plurality of paths between the first node and the one or more third nodes, The one or more third nodes are configured to transmit the encoded data unit to the second node via one or more paths between the one or more third nodes and the second node.

13. The method of claim 12, wherein: The one or more third nodes are configured to decode together data units received from two or more different first nodes.

14. The method of claim 1, further comprising: sending a control frame between the first node and the second node, the control frame including information about the cost; as well as Based on a given maximum target cost, one or more transmission characteristics of the path are adjusted.

15. The method of claim 1, further comprising: Based on the determined costs, at least one of the second paths is closed and / or additional second paths are opened.

16. A method for data transmission, comprising: obtaining, at a first node, a plurality of original packets to be transmitted to a second node; generating, by the first node, a decoded packet by linearly combining two or more of the plurality of original packets using network decoding; determining, by the first node, the second node, or both, a cost associated with transmitting data over each of a plurality of paths; transmitting the decoded packet from the first node to the second node via one or more first paths of the plurality of paths; generating, by the first node, a re-decoded packet by linearly combining two or more of the decoded packets; distributing, by the first node, the re-decoded packets among one or more second paths of the plurality of paths based on the cost; as well as The re-decoded packets are transmitted, by the first node, to the second node via the one or more second paths.

17. A method for data transmission operating in a system having one or more first nodes, at least one second node, and a third node, wherein: The second node and the third node are capable of establishing a transmission session and transmitting encoded data therebetween via one or more available, different output paths between the second node and the third node, the method comprising: transmitting data from the one or more first nodes to the third node via one or more first paths; encoding, by the second node, the data to produce encoded data if the data has not been previously decoded by the network; If the data was previously network-coded, re-coding the data by the second node to produce re-coded data; and The encoded data and / or the re-decoded data are transmitted by the second node from the second node to the third node via one or more second ones of the output paths.

18. An apparatus for use at a first node to transmit data to a second node, the apparatus comprising: processor; as well as a non-volatile memory storing computer program code which, when executed on the processor, causes the processor to perform a process, the process being operable to: transmitting a first data unit from the first node to the second node via a first path; receiving information about one or more second paths between the first node and the second node, the received information comprising information about path availability or path requirements; determining a cost associated with the first path and each of the one or more second paths; transmitting a second data unit from the first node to the second node via the first path; as well as The network-decoded data unit is generated by: If network decoding is not used in the first path, decoding the second data unit, and If network decoding is used in the first path, re-decoding the second data unit; as well as The network coded data unit is transmitted from the first node to the second node via the second path.

19. The apparatus of claim 18, wherein: Receiving information regarding availability of the one or more second paths includes receiving the information in a data request sent by the second node.

20. The apparatus of claim 18, comprising: In response to receiving information regarding availability of one or more second paths between the first node and the second node: generating redundant degrees of freedom DoF by applying network decoding to the second data unit; as well as The redundant DoF is transmitted from the first node to the second node via the first path and the one or more second paths.

21. The apparatus of claim 18, wherein: Determining a cost associated with each of the one or more second paths includes determining one or more cost factors associated with each of the one or more second paths.

22. The apparatus of claim 21, wherein: The one or more cost factors include at least one of the following: the absolute cost factor associated with the path; the relative cost factors of two different paths; or a total cost factor associated with two or more paths, and The one or more cost factors include at least one of the following: Line delays; Cost of use; load; congestion; or Line variability.