A Random Linear Network Coding Approach to Multicast
Summary (3 min read)
Introduction
- This paper is an initial exploration of random linear network coding, posing more questions that it answers.
- Resource consumption can naturally be traded off against capacity and robustness, and across multiple communicating sessions; subsequent work on distributed resource optimization, e.g., [13], [21], has used random linear network coding as a component of the solution.
A. Overview
- In Section II, the authors describe the network model and algebraic coding approach they use in their analyses, and introduce some notation and existing results.
- Section III gives some insights arising from consideration of bipartite matching and network flows.
- Success/error probability bounds for random linear network coding are given for independent and linearly correlated sources in Section IV and for arbitrarily correlated sources in Section V.
- The authors also give examples of practical scenarios in which randomized network coding can be advantageous compared to routing, in Section VI.
- The authors present their conclusions and some directions for further work in Section VII.
A. Basic Model
- Nodes and are called the origin and destination, respectively, of link .
- The authors consider the multicast case where for all .
- For the case of linearly correlated sources, the authors assume that the sources can be modeled as given linear combinations of underlying independent source processes, each with an entropy rate of one bit per unit time, as described further in Section II-B.
- For the latter case, the authors consider general networks without buffering, and make the simplifying assumption that each link has the same delay.
VI. BENEFITS OF RANDOMIZED CODING OVER ROUTING
- Network coding, as a superset of routing, has been shown to offer significant capacity gains for networks with special structure [26].
- For many other networks, network coding does not give higher capacity than centralized optimal routing, but can offer other advantages when centralized optimal routing is difficult.
- The authors consider two types of network scenarios in which distributed random linear coding can be particularly useful.
A. Distributed Settings
- In networks with large numbers of nodes and/or changing topologies, it may be expensive or infeasible to reliably maintain routing state at network nodes.
- The source node sends one process in both directions on one axis and the other process in both directions along the other axis, as illustrated in Fig.
- A node receiving information on two links sends one of the incoming processes on one of its two outgoing links with equal probability, and the other process on the remaining link.
- For the randomized flooding scheme RF, the probability that a receiver located at grid position relative to the source can decode both source processes is at least, also known as 5 Proposition 1.
- 5This simple scheme, unlike the randomized flooding scheme RF, leaves out the optimization that each node receiving two linearly independent processes should always send out two linearly independent processes.
B. Dynamically Varying Connections
- Another scenario in which random linear network coding can be advantageous is for multisource multicast with dynamically varying connections.
- The parameter values for the tests were chosen such that the resulting random graphs would in general be connected and able to support some of the desired connections, while being small enough for the simulations to run efficiently.
- In each time slot, a source was either on, i.e., transmitting source informa- tion, or off, i.e., not transmitting source information.
- Each of these types of overhead depends on the coding field size.
- To this end, the authors use a small field size that allows random linear coding to generally match the performance of the Steiner heuristic, and to surpass it in networks whose topology makes Steiner tree routing difficult.
VII. CONCLUSION
- The authors have presented a distributed random linear network coding approach which asymptotically achieves capacity, as given by the max-flow min-cut bound of [1], in multisource multicast networks.
- These examples suggest that the decentralized nature and robustness of random linear network coding can offer significant advantages in settings that hinder optimal centralized network control.
- Further work includes extensions to nonuniform code distributions, possibly chosen adaptively or with some rudimentary coordination, to optimize different performance goals.
- The randomized and distributed nature of the approach also leads us naturally to consider applications in network security.
- It would also be interesting to consider protocol issues for different communication scenarios, and to compare specific coding and routing protocols over a range of performance metrics.
Did you find this useful? Give us your feedback
Citations
2,442 citations
2,190 citations
1,919 citations
Cites background from "A Random Linear Network Coding Appr..."
...The studies by Ho et al. [ 22 ] and Sanders et al. [23] further showed that random linear network coding over a sufficiently large finite field can (asymptotically) achieve the multicast capacity....
[...]
...Further, simple random linear combinations will suffice with high probability as the field size over which coding is performed grows, as shown by Ho. et al. [ 22 ]....
[...]
1,525 citations
1,389 citations
References
8,533 citations
"A Random Linear Network Coding Appr..." refers background or methods in this paper
...THE capacity of multicast networks with network coding was given in [1]....
[...]
...[1] showed that with network coding, as symbol size approaches infinity, a source can multicast information at a rate approaching the smallest minimum cut between the source and any receiver....
[...]
...Reference [1] shows that coding enables the multicast infor-...
[...]
...given by the max-flow min-cut bound of [1], in multisource...
[...]
...A cyclic graph with nodes and rate may also be converted to an expanded acyclic graph with nodes and rate at least , communication on which can be emulated over time steps on the original cyclic graph [1]....
[...]
4,412 citations
4,165 citations
"A Random Linear Network Coding Appr..." refers background in this paper
...Second, in the context of a distributed source coding problem, we demonstrate that random linear coding also performs compression when necessary in a network, generalizing known error exponents for linear Slepian–Wolf coding [4] in a natural way....
[...]
...The error exponents for general networks reduce to those obtained in [4] for the Slepian–Wolf network where , , , TABLE I SUCCESS PROBABILITIES OF RANDOMIZED FLOODING SCHEME RF AND RANDOM LINEAR CODING SCHEME RC....
[...]
...Analogously to Slepian and Wolf [28], we consider the problem of distributed encoding and joint decoding of two sources whose output values in each unit time period are drawn independent and identically distributed (i....
[...]
...Analogously to Slepian and Wolf [28], we consider the problem of distributed encoding and joint decoding of two sources whose output values in each unit time period are drawn independent and identically distributed (i.i.d.) from the same joint distribution ....
[...]
...In the special case of a network consisting of one direct link from each source to a common receiver, this reduces to the original Slepian–Wolf problem....
[...]
3,660 citations
2,628 citations
"A Random Linear Network Coding Appr..." refers background or methods in this paper
...We used these formulations in obtaining a tighter upper bound on the required field size than the previous bound of [17], and in our analysis of distributed randomized network coding, introduced in [9]....
[...]
...a receiver is given by the transfer matrix [17]....
[...]
...Note, however, that the linear decoding strategies of [17] do not apply for the case of arbitrarily correlated sources....
[...]
...In the scalar algebraic coding framework of [17], the source information processes, the receiver output processes, and the in-...
[...]
...The need for vector coding solutions in some nonmulticast problems was considered by Rasala Lehman and Lehman [18], Médard et al. [22], and Riis [25]....
[...]
Related Papers (5)
Frequently Asked Questions (12)
Q2. What have the authors stated for future works in "A random linear network coding approach to multicast" ?
Further work includes extensions to nonuniform code distributions, possibly chosen adaptively or with some rudimentary coordination, to optimize different performance goals.
Q3. What is the way to get the upper bounds on coding field size?
Lower bounds on coding field size were presented by Rasala Lehman and Lehman [18] and Feder et al. [6]. [6] also gave graph-specific upper bounds based on the number of “clashes” between flows from source to terminals.
Q4. What is the probability that a random network code is valid for the problem?
If there exists a solution to the network connection problem with the same values for the fixed code coefficients, then the probability that the random network code is valid for the problem is at least , where is the maximum number of links with associated random coefficients in any set of links constituting a flow solution for any receiver.
Q5. What is the probability of a network connection problem where the code coefficients are fixed?
Nodes that cannot determine the appropriate code coefficients from local information choose the coefficients independently and uniformly from .
Q6. What is the general bound on the success probability of random linear network coding?
The authors have given a general bound on the success probability of such codes for arbitrary networks, showing that error probability decreases exponentially with code length.
Q7. What is the probability that the random network code is valid for the problem?
If there exists a solution to the network connection problem with the same values for the fixed code coefficients, then the probability that the random network code is valid for the problem is at least , where is the number of links with associated random coefficients.
Q8. What are the simplest assumptions for the case of arbitrarily correlated sources?
For the case of arbitrarily correlated sources, the authors consider sources with integer bit rates and arbitrary joint probability distributions.
Q9. What is the next bound for the acyclic delay-free case?
The next bound is useful in cases where analysis of connection feasibility is easier than direct analysis of random linear coding.
Q10. What is the way to achieve the performance of the Steiner tree?
To this end, the authors use a small field size that allows random linear coding to generally match the performance of the Steiner heuristic, and to surpass it in networks whose topology makes Steiner tree routing difficult.
Q11. What is the probability of a randomlinear code being valid?
it is intuitive that having more redundant capacity in the network, for instance, should increase the probability that a randomlinear code will be valid.
Q12. What are the advantages of random linear network coding?
These examples suggest that the decentralized nature and robustness of random linear network coding can offer significant advantages in settings that hinder optimal centralized network control.