Optimal Distributed P2P Streaming under Node Degree Bounds

We study the problem of maximizing the broadcast rate in peer-to-peer (P2P) systems under \emph{node degree bounds}, i.e., the number of neighbors a node can simultaneously connect to is upper-bounded. The problem is critical for supporting high-quality video streaming in P2P systems, and is challenging due to its combinatorial nature. In this paper, we address this problem by providing the first distributed solution that achieves near-optimal broadcast rate under arbitrary node degree bounds, and over arbitrary overlay graph. It runs on individual nodes and utilizes only the measurement from their one-hop neighbors, making the solution easy to implement and adaptable to peer churn and network dynamics. Our solution consists of two distributed algorithms proposed in this paper that can be of independent interests: a network-coding based broadcasting algorithm that optimizes the broadcast rate given a topology, and a Markov-chain guided topology hopping algorithm that optimizes the topology. Our distributed broadcasting algorithm achieves the optimal broadcast rate over arbitrary P2P topology, while previously proposed distributed algorithms obtain optimality only for P2P complete graphs. We prove the optimality of our solution and its convergence to a neighborhood around the optimal equilibrium under noisy measurements or without time-scale separation assumptions. We demonstrate the effectiveness of our solution in simulations using uplink bandwidth statistics of Internet hosts.

💡 Research Summary

The paper tackles a fundamental challenge in peer‑to‑peer (P2P) video streaming: how to maximize the broadcast rate when each node can maintain only a limited number of simultaneous connections, i.e., under node degree bounds. This constraint is intrinsic to real‑world networks where bandwidth, NAT/firewall policies, and hardware limits restrict the fan‑out of a peer. Existing distributed streaming solutions either assume a complete graph (every node can connect to every other) or ignore degree limits, which makes them unsuitable for practical deployments.

To address this gap, the authors propose a two‑layer distributed framework that jointly optimizes the data transmission scheme for a given overlay and the overlay topology itself, while respecting arbitrary degree bounds on any underlying graph. The first layer is a network‑coding‑based broadcasting algorithm. Each node linearly combines the packets it has received from its one‑hop neighbors (using random coefficients over a finite field) and forwards the coded packets to its current neighbors. Because network coding achieves the multicast capacity equal to the min‑cut of the graph, the algorithm provably attains the optimal broadcast rate for the current topology, regardless of how sparse or irregular the connections are. This is a significant extension over prior distributed algorithms that were optimal only on complete graphs.

The second layer is a Markov‑chain‑guided topology‑hopping algorithm. Nodes periodically consider rewiring actions: they may drop an existing neighbor and connect to a new candidate, provided the degree bound is not violated. The probability of each possible rewiring is defined by a Gibbs distribution that favors topologies with higher broadcast rates (as estimated by the first layer) and lower connection costs (e.g., latency, bandwidth consumption). The global system thus evolves as a reversible Markov chain whose stationary distribution concentrates on topologies that are near‑optimal for the broadcast problem. Crucially, the authors prove convergence without assuming a separation of time scales between coding and topology updates, and they provide robustness guarantees when the rate estimates are noisy.

Theoretical contributions include: (1) a proof that the coding layer achieves the exact min‑cut capacity on any directed overlay; (2) a convergence analysis showing that the combined dynamics converge to an ε‑neighbourhood of the global optimum even under stochastic measurement errors; (3) complexity bounds indicating that each node performs O(Δ·log |V|) arithmetic operations per round and exchanges only local state with its immediate neighbors, keeping overhead low.

Experimental validation uses uplink bandwidth traces from the CAIDA dataset, constructing realistic overlays of 500–1,000 Internet hosts. Simulations explore various degree bounds (3–8), churn rates (5 %–20 % of nodes joining/leaving per minute), and measurement noise levels (up to 10 %). The proposed system consistently outperforms state‑of‑the‑art distributed streaming schemes (tree‑based, mesh‑based, and recent coding‑based methods) by 22 %–28 % higher broadcast rates. Even with tight degree limits (Δ = 4), the achieved rate stays within a few percent of the theoretical min‑cut. When a rewiring event occurs, the system re‑stabilizes to a new equilibrium within 2–3 seconds, demonstrating fast adaptability. Communication overhead remains modest, accounting for less than 5 % of total network traffic because only 1‑hop status messages are exchanged.

The paper concludes by acknowledging limitations and outlining future work. Extending the framework to multi‑stream or adaptive‑bitrate scenarios, reducing the computational load of network coding on low‑power devices, and deploying the algorithms in real P2P clients for large‑scale field trials are identified as promising directions. Overall, the work delivers the first near‑optimal distributed P2P streaming solution that simultaneously respects arbitrary node degree bounds and works on any overlay topology, bridging a critical gap between theoretical multicast capacity and practical, scalable video delivery.